25_downsample_cust_hier.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "8816b366-5be4-4748-b769-4dc6f7d1735f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Modify the file 00_setup.py to define input/output file paths on your system\n",
    "# The information in 00_setup.py will be used across notebooks\n",
    "from importlib.machinery import SourceFileLoader\n",
    "setup = SourceFileLoader(\"setup\", \"./00_setup.py\").load_module()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a945de0c-f042-46b2-9d06-ab064af6c05b",
   "metadata": {},
   "source": [
    "## NN, Downsample train, NAICS only, Data Generator Randomize\n",
    "In the downsampled data, use the custom data generator to inject 20% random cases for each batch.  This is to help the model learn to handle unseen codes.  But downsample the training data to examine the "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9e6fba8f-d564-466c-8e50-7773b095157f",
   "metadata": {},
   "source": [
    "Use higher levels of the NAICS hierarchy, and data with no missing values"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "32221be3-556a-441d-a30c-1f22f5dd664f",
   "metadata": {},
   "source": [
    "Custom data generators are used to retrieve data at each batch or epoch, and can be used to modify data for every batch or epoch.  In future scripts, I modify at the batch level, injecting random \"1\" values.   See also  https://www.tensorflow.org/versions/r2.1/api_docs/python/tf/keras/utils/Sequence"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d47a7ec9-7d7b-47df-a902-b2f1e54f1135",
   "metadata": {},
   "source": [
    "*This script takes about 15 minutes on my MacBook Air*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "fb42f9b4-99a8-4753-aacf-8722469be4c8",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import tensorflow as tf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "bf51170a-4487-4d10-b63d-c88f41b529f9",
   "metadata": {},
   "outputs": [],
   "source": [
    "from pathlib import Path\n",
    "import importlib, pickle\n",
    "from sklearn import model_selection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "09dd973b-60a0-434b-9193-8c5fcec27dd5",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import os\n",
    "import re\n",
    "\n",
    "import keras\n",
    "from tensorflow.keras import layers, optimizers, losses, metrics, Model\n",
    "from sklearn import preprocessing, feature_extraction, model_selection\n",
    "from IPython.display import display, HTML\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "5dedd82d-1901-4327-8b7f-f00026a84f18",
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.callbacks import EarlyStopping"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "9f1e4f2e-0e45-4e28-8e9c-212e0bb51199",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sba_nn.sba_nn import sbnn_metrics, sbnn_model\n",
    "from sba_nn.sba_nn.sbnn_model import CatInjectGenerator"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3cbc00d2-bc83-42d9-83eb-65fcf306dccd",
   "metadata": {},
   "source": [
    "## Input Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "e014cd31-0f8d-42e3-82f9-b529eec8a3cc",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "sba_loans = pd.read_parquet(Path(setup.temp_path).joinpath('20_DATA_downsamp_train.parquet'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "8e65cb7c-792c-46ff-a530-0d5c64240d60",
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(Path(setup.temp_path).joinpath('01_DATA_features.pkl'), 'rb') as fin:\n",
    "    imputer_features = pickle.load(fin)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "6f0ce28e-a80c-415b-b068-5c4d49370da8",
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(Path(setup.temp_path).joinpath('01_DATA_naics_max_encodings.pkl'), 'rb') as fin:\n",
    "    naics_max_levels= pickle.load(fin)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4d4116da-d689-478a-9c5d-a330f231e615",
   "metadata": {},
   "source": [
    "##### Features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "f98c6e41-38d0-4d5f-b821-e3837b8c2068",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['NoEmp', 'CreateJob', 'LowDoc', 'DisbursementGross', 'new_business', 'urban_flag', 'franchise_flag', 'missingindicator_LowDoc', 'missingindicator_new_business', 'missingindicator_urban_flag']\n"
     ]
    }
   ],
   "source": [
    "# Numeric features\n",
    "features_numeric = [f for f in imputer_features if 'NAICS' not in f]\n",
    "features_numeric_len = len(features_numeric)\n",
    "print(features_numeric)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "a74fa386-4011-45ee-bdb5-d671db9ffa91",
   "metadata": {},
   "outputs": [],
   "source": [
    "# NAICS feature\n",
    "features_naics = ['NAICS', 'NAICS_4', 'NAICS_3', 'NAICS_sector']\n",
    "features_naics_max_levels  = [naics_max_levels[n] for n in features_naics]\n",
    "features_naics_emb_dim = [setup.nn_naics_embed_size_dict[n] for n in features_naics]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "867031ee-8b6e-424d-b497-8e19409ba1d9",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1170, 345, 107, 21]\n",
      "[8, 4, 4, 2]\n"
     ]
    }
   ],
   "source": [
    "print(features_naics_max_levels)\n",
    "print(features_naics_emb_dim)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b979396c-c822-4940-a292-6e07f1b54f58",
   "metadata": {},
   "source": [
    "##### Datasets for train, validation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "43bca011-a59d-4496-9f03-87f4e350a253",
   "metadata": {},
   "outputs": [],
   "source": [
    "X = sba_loans[['dset', 'LoanNr_ChkDgt'] + features_numeric + features_naics].set_index('LoanNr_ChkDgt').sort_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "aa162d01-ed73-4ae3-832e-4aeea09b70f4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "training X: (44684, 14), y:(44684,)\n"
     ]
    }
   ],
   "source": [
    "X_train = X[X['dset'] == 'train'].drop(columns='dset')\n",
    "y_train = sba_loans[sba_loans['dset'] == 'train'].set_index('LoanNr_ChkDgt').sort_index()['target']\n",
    "print(f'training X: {X_train.shape}, y:{y_train.shape}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "b32b5fd3-6f6e-4aec-8e20-0e2b71a4814c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "val X: (95604, 14), y:(95604,)\n"
     ]
    }
   ],
   "source": [
    "X_val = X[X['dset'] == 'val'].drop(columns='dset')\n",
    "y_val = sba_loans[sba_loans['dset'] == 'val'].set_index('LoanNr_ChkDgt').sort_index()['target']\n",
    "print(f'val X: {X_val.shape}, y:{y_val.shape}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "01b8ff96-159b-4283-b5fb-c4b5dbd277d5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.20295855339718916\n"
     ]
    }
   ],
   "source": [
    "base_thresh = y_train.mean()\n",
    "print(base_thresh)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3db4ea1c-77b2-4721-b106-1e66fc502139",
   "metadata": {},
   "source": [
    "## Create, fit model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "4d48ddfa-580c-4321-9917-0bd176b67d87",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-05-20 13:34:09.106744: I metal_plugin/src/device/metal_device.cc:1154] Metal device set to: Apple M1\n",
      "2024-05-20 13:34:09.106797: I metal_plugin/src/device/metal_device.cc:296] systemMemory: 16.00 GB\n",
      "2024-05-20 13:34:09.106811: I metal_plugin/src/device/metal_device.cc:313] maxCacheSize: 5.33 GB\n",
      "2024-05-20 13:34:09.106886: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:303] Could not identify NUMA node of platform GPU ID 0, defaulting to 0. Your kernel may not have been built with NUMA support.\n",
      "2024-05-20 13:34:09.106921: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:269] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 0 MB memory) -> physical PluggableDevice (device: 0, name: METAL, pci bus id: <undefined>)\n"
     ]
    }
   ],
   "source": [
    "this_model = sbnn_model.create_emb_model(features_numeric_len,\n",
    "                                         features_naics_max_levels,\n",
    "                                         features_naics_emb_dim,\n",
    "                                         naics_embedding_names = features_naics,\n",
    "                                         hidden_size = setup.nn_layer_sizes,\n",
    "                                         activation='tanh',\n",
    "                                         lr=setup.nn_learning_rate,\n",
    "                                         opt_func = setup.nn_optimizer,\n",
    "                                         dropout = setup.nn_dropout)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "47036621-ef13-4957-997b-f8ef86016f7f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "__________________________________________________________________________________________________\n",
      " Layer (type)                Output Shape                 Param #   Connected to                  \n",
      "==================================================================================================\n",
      " input_naics_00 (InputLayer  [(None, 1)]                  0         []                            \n",
      " )                                                                                                \n",
      "                                                                                                  \n",
      " input_naics_01 (InputLayer  [(None, 1)]                  0         []                            \n",
      " )                                                                                                \n",
      "                                                                                                  \n",
      " input_naics_02 (InputLayer  [(None, 1)]                  0         []                            \n",
      " )                                                                                                \n",
      "                                                                                                  \n",
      " input_naics_03 (InputLayer  [(None, 1)]                  0         []                            \n",
      " )                                                                                                \n",
      "                                                                                                  \n",
      " emb_naics_00 (Embedding)    (None, 1, 8)                 9360      ['input_naics_00[0][0]']      \n",
      "                                                                                                  \n",
      " emb_naics_01 (Embedding)    (None, 1, 4)                 1380      ['input_naics_01[0][0]']      \n",
      "                                                                                                  \n",
      " emb_naics_02 (Embedding)    (None, 1, 4)                 428       ['input_naics_02[0][0]']      \n",
      "                                                                                                  \n",
      " emb_naics_03 (Embedding)    (None, 1, 2)                 42        ['input_naics_03[0][0]']      \n",
      "                                                                                                  \n",
      " NAICS (Flatten)             (None, 8)                    0         ['emb_naics_00[0][0]']        \n",
      "                                                                                                  \n",
      " NAICS_4 (Flatten)           (None, 4)                    0         ['emb_naics_01[0][0]']        \n",
      "                                                                                                  \n",
      " NAICS_3 (Flatten)           (None, 4)                    0         ['emb_naics_02[0][0]']        \n",
      "                                                                                                  \n",
      " NAICS_sector (Flatten)      (None, 2)                    0         ['emb_naics_03[0][0]']        \n",
      "                                                                                                  \n",
      " input_numeric (InputLayer)  [(None, 10)]                 0         []                            \n",
      "                                                                                                  \n",
      " input_concat (Concatenate)  (None, 28)                   0         ['NAICS[0][0]',               \n",
      "                                                                     'NAICS_4[0][0]',             \n",
      "                                                                     'NAICS_3[0][0]',             \n",
      "                                                                     'NAICS_sector[0][0]',        \n",
      "                                                                     'input_numeric[0][0]']       \n",
      "                                                                                                  \n",
      " layer_00 (Dense)            (None, 128)                  3712      ['input_concat[0][0]']        \n",
      "                                                                                                  \n",
      " dropout_00 (Dropout)        (None, 128)                  0         ['layer_00[0][0]']            \n",
      "                                                                                                  \n",
      " layer_01 (Dense)            (None, 64)                   8256      ['dropout_00[0][0]']          \n",
      "                                                                                                  \n",
      " dropout_01 (Dropout)        (None, 64)                   0         ['layer_01[0][0]']            \n",
      "                                                                                                  \n",
      " output (Dense)              (None, 1)                    65        ['dropout_01[0][0]']          \n",
      "                                                                                                  \n",
      "==================================================================================================\n",
      "Total params: 23243 (90.79 KB)\n",
      "Trainable params: 23243 (90.79 KB)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "this_model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4ed7408d-99ce-413e-aef8-23743b5601cd",
   "metadata": {},
   "source": [
    "##### Use data generator to inject 1 values into training data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "f1dfde86-af9b-4110-aae9-30f3d59404af",
   "metadata": {},
   "outputs": [],
   "source": [
    "generator = CatInjectGenerator(X_train, y_train,\n",
    "                               categorical_columns=features_naics,\n",
    "                               batch_size = setup.nn_batch_size,\n",
    "                               injection_rate = 0.2,\n",
    "                               injection_value = 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "dceea363-56c3-4ef7-a0fa-f9ff3e6845a5",
   "metadata": {},
   "outputs": [],
   "source": [
    "callback = keras.callbacks.EarlyStopping(patience=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "fd9ebbdb-daae-45c7-bbe2-751f8c245c31",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-05-20 13:34:11.408131: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:114] Plugin optimizer for device_type GPU is enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1397/1397 [==============================] - ETA: 0s - loss: 0.4843 - auc: 0.2913 - auc_1: 0.6554"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-05-20 13:35:03.413288: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:114] Plugin optimizer for device_type GPU is enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1397/1397 [==============================] - 94s 65ms/step - loss: 0.4843 - auc: 0.2913 - auc_1: 0.6554 - val_loss: 0.4639 - val_auc: 0.3428 - val_auc_1: 0.7038\n",
      "Epoch 2/20\n",
      "1397/1397 [==============================] - 87s 63ms/step - loss: 0.4662 - auc: 0.3342 - auc_1: 0.6952 - val_loss: 0.4610 - val_auc: 0.3476 - val_auc_1: 0.7074\n",
      "Epoch 3/20\n",
      "1397/1397 [==============================] - 85s 61ms/step - loss: 0.4618 - auc: 0.3441 - auc_1: 0.7044 - val_loss: 0.4627 - val_auc: 0.3457 - val_auc_1: 0.7047\n",
      "Epoch 4/20\n",
      "1397/1397 [==============================] - 85s 61ms/step - loss: 0.4596 - auc: 0.3478 - auc_1: 0.7096 - val_loss: 0.4606 - val_auc: 0.3524 - val_auc_1: 0.7090\n",
      "Epoch 5/20\n",
      "1397/1397 [==============================] - 86s 62ms/step - loss: 0.4587 - auc: 0.3499 - auc_1: 0.7103 - val_loss: 0.4603 - val_auc: 0.3544 - val_auc_1: 0.7098\n",
      "Epoch 6/20\n",
      "1397/1397 [==============================] - 77s 55ms/step - loss: 0.4560 - auc: 0.3551 - auc_1: 0.7157 - val_loss: 0.4615 - val_auc: 0.3559 - val_auc_1: 0.7100\n",
      "Epoch 7/20\n",
      "1397/1397 [==============================] - 55s 39ms/step - loss: 0.4542 - auc: 0.3601 - auc_1: 0.7198 - val_loss: 0.4600 - val_auc: 0.3555 - val_auc_1: 0.7101\n",
      "Epoch 8/20\n",
      "1397/1397 [==============================] - 45s 32ms/step - loss: 0.4525 - auc: 0.3649 - auc_1: 0.7228 - val_loss: 0.4634 - val_auc: 0.3518 - val_auc_1: 0.7069\n",
      "Epoch 9/20\n",
      "1397/1397 [==============================] - 41s 29ms/step - loss: 0.4521 - auc: 0.3661 - auc_1: 0.7234 - val_loss: 0.4608 - val_auc: 0.3552 - val_auc_1: 0.7105\n",
      "Epoch 10/20\n",
      "1397/1397 [==============================] - 41s 29ms/step - loss: 0.4515 - auc: 0.3687 - auc_1: 0.7245 - val_loss: 0.4605 - val_auc: 0.3576 - val_auc_1: 0.7116\n"
     ]
    }
   ],
   "source": [
    "this_history = this_model.fit(generator,\n",
    "                              batch_size=setup.nn_batch_size,\n",
    "                              epochs = setup.nn_epochs,\n",
    "                              validation_data=([X_val[features_numeric]] + [X_val[f] for f in features_naics], \n",
    "                                               y_val),\n",
    "                              shuffle=True,\n",
    "                             callbacks=[callback])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "460c4b43-c673-4ee1-b070-d2d596ac4561",
   "metadata": {},
   "outputs": [],
   "source": [
    "this_history_df = sbnn_model.process_history(this_history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "ad4783f2-8d0b-4850-a45a-3a4be8e63318",
   "metadata": {},
   "outputs": [],
   "source": [
    "this_history_df.to_csv(Path(setup.temp_path).joinpath('25_REPORT_fit_history.csv'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "73e89efd-75a9-4b27-81a8-13e12d5482be",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Axes: >"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjUAAAGdCAYAAADqsoKGAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/OQEPoAAAACXBIWXMAAA9hAAAPYQGoP6dpAABgoElEQVR4nO3deXhU9b3H8ffMZA9ZCCEbBMKeIHuACAKKRMEFRahVi4KIYK1oJdUq3lYLvRZtK+XWDaVgXQt1rYpSNcoSiKBgAFkCBMKekASyQ7aZ+8dJJsQMykDCJDOf1/PMQ+bMOTPfIZD55LeabDabDREREZFWzuzqAkRERESagkKNiIiIuAWFGhEREXELCjUiIiLiFhRqRERExC0o1IiIiIhbUKgRERERt6BQIyIiIm7By9UFXCxWq5WjR48SFBSEyWRydTkiIiJyDmw2GyUlJcTExGA2/3hbjMeEmqNHjxIbG+vqMkREROQ8HDp0iI4dO/7oOR4TaoKCggDjLyU4ONjF1YiIiMi5KC4uJjY21v45/mM8JtTUdTkFBwcr1IiIiLQy5zJ0RAOFRURExC0o1IiIiIhbOK9Q8/zzzxMXF4efnx9JSUls3LjxnK5btmwZJpOJCRMmNDheWlrKrFmz6NixI/7+/vTu3ZtFixY1OOeKK67AZDI1uP3yl788n/JFRETEDTk9pmb58uWkpKSwaNEikpKSWLhwIWPHjiUzM5OIiIizXpednc1DDz3EyJEjGz2WkpLCl19+yRtvvEFcXByfffYZv/rVr4iJieGGG26wnzdjxgzmzZtnvx8QEOBs+SIi4sFsNhvV1dXU1NS4uhSpZbFY8PLyapLlVpwONQsWLGDGjBlMmzYNgEWLFrFixQqWLl3Ko48+6vCampoaJk+ezNy5c1m7di2FhYUNHl+/fj1Tp07liiuuAGDmzJm89NJLbNy4sUGoCQgIICoqytmSRUREqKys5NixY5SXl7u6FPmBgIAAoqOj8fHxuaDncSrUVFZWsmnTJubMmWM/ZjabSU5OJj09/azXzZs3j4iICKZPn87atWsbPT58+HA+/PBD7rrrLmJiYli1ahW7d+/mb3/7W4Pz3nzzTd544w2ioqIYP348v//978/aWlNRUUFFRYX9fnFxsTNvVURE3IjVamX//v1YLBZiYmLw8fHRQqwtgM1mo7Kykry8PPbv30+PHj1+coG9H+NUqMnPz6empobIyMgGxyMjI9m1a5fDa9LS0liyZAkZGRlnfd5nn32WmTNn0rFjR7y8vDCbzSxevJhRo0bZz/nFL35B586diYmJYevWrTzyyCNkZmby3nvvOXzO+fPnM3fuXGfenoiIuKnKykqsViuxsbEautDC+Pv74+3tzYEDB6isrMTPz++8n6tZ16kpKSnhjjvuYPHixYSHh5/1vGeffZavv/6aDz/8kM6dO7NmzRruu+8+YmJiSE5OBowuqTp9+/YlOjqaMWPGkJWVRbdu3Ro955w5c0hJSbHfr1u8R0REPNeFtAJI82mq74tToSY8PByLxUJubm6D47m5uQ7HumRlZZGdnc348ePtx6xWq/HCXl5kZmYSExPDY489xvvvv891110HQL9+/cjIyOCvf/2rPdT8UFJSEgB79+51GGp8fX3x9fV15u2JiIhIK+ZUNPLx8SExMZHU1FT7MavVSmpqKsOGDWt0fnx8PNu2bSMjI8N+u+GGGxg9ejQZGRnExsZSVVVFVVVVo5RmsVjsAciRuu6s6OhoZ96CiIiIuCmnu59SUlKYOnUqgwcPZujQoSxcuJCysjL7bKgpU6bQoUMH5s+fj5+fH3369GlwfWhoKID9uI+PD5dffjkPP/ww/v7+dO7cmdWrV/Paa6+xYMECwGjxeeutt7j22mtp164dW7duZfbs2YwaNYp+/fpdyPsXERFp0a644goGDBjAwoULXV1Ki+d0qLnlllvIy8vj8ccfJycnhwEDBrBy5Ur74OGDBw863Te2bNky5syZw+TJkzlx4gSdO3fmySeftC+u5+PjwxdffGEPULGxsUyaNInf/e53zpYvIiIibspks9lsri7iYiguLiYkJISioqIm3dBy7/ES3tl0hLYB3txzeeOxPSIi4nqnT59m//79dOnS5YJm17iCJ7TU/Nj3x5nPbw0Dv0D78spYtDqLZd8ccnUpIiJyjmw2G+WV1S65XUhbwsmTJ5kyZQpt27YlICCAa665hj179tgfP3DgAOPHj6dt27YEBgZyySWX8Mknn9ivnTx5Mu3bt8ff358ePXrwyiuvXPDfZUvSrFO6PcGl3dphMZvYn1/GoRPlxIZp/QMRkZbuVFUNvR//r0tee8e8sQT4nN/H75133smePXv48MMPCQ4O5pFHHuHaa69lx44deHt7c99991FZWcmaNWsIDAxkx44dtGnTBoDf//737Nixg08//ZTw8HD27t3LqVOnmvKtuZxCzQUK9vNmYGwo3x44SdrefG4b2snVJYmIiBuqCzPr1q1j+PDhgLHSfmxsLB988AE333wzBw8eZNKkSfTt2xeArl272q8/ePAgAwcOZPDgwQDExcVd9PfQ3BRqmsCIHuF8e+Aka/fkKdSIiLQC/t4Wdswb67LXPh87d+7Ey8vLvk4bQLt27ejVqxc7d+4E4IEHHuDee+/ls88+Izk5mUmTJtlnCd97771MmjSJzZs3c/XVVzNhwgR7OHIXGlPTBEb2MFZLXre3gBqrR4y7FhFp1UwmEwE+Xi65NeeeU3fffTf79u3jjjvuYNu2bQwePJhnn30WgGuuuYYDBw4we/Zsjh49ypgxY3jooYearRZXUKhpAv07hhLk60XRqSq+P1Lk6nJERMQNJSQkUF1dzYYNG+zHCgoKyMzMpHfv3vZjsbGx/PKXv+S9997jN7/5DYsXL7Y/1r59e6ZOncobb7zBwoULefnlly/qe2huCjVNwMtiZli3dgCs3ZPn4mpERMQd9ejRgxtvvJEZM2aQlpbGli1buP322+nQoQM33ngjAA8++CD//e9/2b9/P5s3b+arr74iISEBgMcff5z//Oc/7N27l+3bt/Pxxx/bH3MXCjVNZGTP9gCs3ZPv4kpERMRdvfLKKyQmJnL99dczbNgwbDYbn3zyCd7e3gDU1NRw3333kZCQwLhx4+jZsycvvPACYCxkO2fOHPr168eoUaOwWCwsW7bMlW+nyWnxvSaSnV/GFX9dhbfFRMbjVxPoqzHYIiItRWtefM8TaPG9FqZzuwBiw/ypqrGxYX+Bq8sRERHxOAo1TcRkMjGiu9EFtWa3uqBEREQuNoWaJjSqdmp32l6FGhERkYtNoaYJDe8WjtkEe4+XcqzIvZaeFhERaekUappQSIA3/TqGApoFJSIicrEp1DSxutWFFWpEREQuLoWaJjayhzFYeN3efKzaMkFEROSiUahpYgM7hRLoY+FEWSU7jhW7uhwRERGPoVDTxLwtZi7tWrdlgrqgRERELhaFmmYw0j61W/tAiYiIa8XFxbFw4cJzOtdkMvHBBx80az3NSaGmGYyoHVfzzf6TnKqscXE1IiIinkGhphl0ax9ITIgflTVWNmafcHU5IiIiHkGhphmYTCZG1E3t3q0uKBGRFsdmg8oy19yc2Ef65ZdfJiYmBqvV2uD4jTfeyF133UVWVhY33ngjkZGRtGnThiFDhvDFF1802V/Ttm3buPLKK/H396ddu3bMnDmT0tJS++OrVq1i6NChBAYGEhoaymWXXcaBAwcA2LJlC6NHjyYoKIjg4GASExP59ttvm6w2R7SVdDMZ2aM9//72sLZMEBFpiarK4U8xrnntx46CT+A5nXrzzTdz//3389VXXzFmzBgATpw4wcqVK/nkk08oLS3l2muv5cknn8TX15fXXnuN8ePHk5mZSadOnS6ozLKyMsaOHcuwYcP45ptvOH78OHfffTezZs3in//8J9XV1UyYMIEZM2bwr3/9i8rKSjZu3IjJZAJg8uTJDBw4kBdffBGLxUJGRgbe3t4XVNNPUahpJpd1D8dkgl05JRwvPk1EsLa6FxER57Rt25ZrrrmGt956yx5q3nnnHcLDwxk9ejRms5n+/fvbz//jH//I+++/z4cffsisWbMu6LXfeustTp8+zWuvvUZgoBHCnnvuOcaPH8/TTz+Nt7c3RUVFXH/99XTr1g2AhIQE+/UHDx7k4YcfJj4+HoAePXpcUD3nQqGmmYQF+tAnJoRtR4pI25vPxEEdXV2SiIjU8Q4wWkxc9dpOmDx5MjNmzOCFF17A19eXN998k1tvvRWz2UxpaSl/+MMfWLFiBceOHaO6uppTp05x8ODBCy5z586d9O/f3x5oAC677DKsViuZmZmMGjWKO++8k7Fjx3LVVVeRnJzMz3/+c6KjowFISUnh7rvv5vXXXyc5OZmbb77ZHn6ai8bUNKMR2jJBRKRlMpmMLiBX3Gq7Z87V+PHjsdlsrFixgkOHDrF27VomT54MwEMPPcT777/Pn/70J9auXUtGRgZ9+/alsrKyOf7WGnnllVdIT09n+PDhLF++nJ49e/L1118D8Ic//IHt27dz3XXX8eWXX9K7d2/ef//9Zq1HoaYZnbkPlM2JgWEiIiJ1/Pz8mDhxIm+++Sb/+te/6NWrF4MGDQJg3bp13Hnnndx000307duXqKgosrOzm+R1ExIS2LJlC2VlZfZj69atw2w206tXL/uxgQMHMmfOHNavX0+fPn1466237I/17NmT2bNn89lnnzFx4kReeeWVJqntbBRqmlFi57b4e1vIL61gV06Jq8sREZFWavLkyaxYsYKlS5faW2nAGKfy3nvvkZGRwZYtW/jFL37RaKbUhbymn58fU6dO5fvvv+err77i/vvv54477iAyMpL9+/czZ84c0tPTOXDgAJ999hl79uwhISGBU6dOMWvWLFatWsWBAwdYt24d33zzTYMxN81BY2qaka+XhaSuYazKzCNtTz4J0cGuLklERFqhK6+8krCwMDIzM/nFL35hP75gwQLuuusuhg8fTnh4OI888gjFxU2z72BAQAD//e9/+fWvf82QIUMICAhg0qRJLFiwwP74rl27ePXVVykoKCA6Opr77ruPe+65h+rqagoKCpgyZQq5ubmEh4czceJE5s6d2yS1nY3J5iH9IsXFxYSEhFBUVERw8MULF/9Yu4//XbGTUT3b89pdQy/a64qISL3Tp0+zf/9+unTpgp+fZqO2ND/2/XHm81vdT81sZO2WCRv2FXC6SlsmiIiINBeFmmbWM7INEUG+VFRb2XTgpKvLERERD/Xmm2/Spk0bh7dLLrnE1eU1CY2paWZ1Wya8t/kIa/bkcVn3cFeXJCIiHuiGG24gKSnJ4WPNvdLvxaJQcxGM6tGe9zYfIW1PPlzj6mpERMQTBQUFERQU5OoympW6ny6CutaZ7UeLyS+tcHE1IiKey0PmxrQ6TfV9Uai5CNoH+dqnc6/TBpciIhddXfdKeXm5iysRR+q+LxfaDabup4tkZI9wdh4rZu2efG4c0MHV5YiIeBSLxUJoaCjHjx8HjDVWTE5uVyBNz2azUV5ezvHjxwkNDcVisVzQ8ynUXCQje4Tz8pp9pNVumaD/TCIiF1dUVBSAPdhIyxEaGmr//lwIhZqLZEhcGD5eZnKKT7P3eCk9It17sJaISEtjMpmIjo4mIiKCqqoqV5cjtby9vS+4haaOQs1F4udtIalLGGv35LN2T75CjYiIi1gslib7EJWW5bwGCj///PPExcXh5+dHUlISGzduPKfrli1bhslkYsKECQ2Ol5aWMmvWLDp27Ii/vz+9e/dm0aJFDc45ffo09913H+3ataNNmzZMmjSJ3Nzc8ynfZUZ0r9u1O8/FlYiIiLgfp0PN8uXLSUlJ4YknnmDz5s3079+fsWPH/mQfZXZ2Ng899BAjR45s9FhKSgorV67kjTfeYOfOnTz44IPMmjWLDz/80H7O7Nmz+eijj3j77bdZvXo1R48eZeLEic6W71L2LRP2n6Cyuml2URURERGD06FmwYIFzJgxg2nTptlbVAICAli6dOlZr6mpqWHy5MnMnTuXrl27Nnp8/fr1TJ06lSuuuIK4uDhmzpxJ//797S1ARUVFLFmyhAULFnDllVeSmJjIK6+8wvr16/n666+dfQsuEx8VRHgbH8ora9h8UFsmiIiINCWnQk1lZSWbNm0iOTm5/gnMZpKTk0lPTz/rdfPmzSMiIoLp06c7fHz48OF8+OGHHDlyBJvNxldffcXu3bu5+uqrAdi0aRNVVVUNXjc+Pp5OnTqd9XUrKiooLi5ucHM1s9mkLigREZFm4lSoyc/Pp6amhsjIyAbHIyMjycnJcXhNWloaS5YsYfHixWd93meffZbevXvTsWNHfHx8GDduHM8//zyjRo0CICcnBx8fH0JDQ8/5defPn09ISIj9Fhsb68Q7bT4jarug0vZoET4REZGm1KwrCpeUlHDHHXewePFiwsPPvpHjs88+y9dff82HH37Ipk2beOaZZ7jvvvv44osvzvu158yZQ1FRkf126NCh836uplTXUrP1SBEnyypdXI2IiIj7cGpKd3h4OBaLpdGso9zcXIeL5mRlZZGdnc348ePtx6xWY4Csl5cXmZmZxMTE8Nhjj/H+++9z3XXXAdCvXz8yMjL461//SnJyMlFRUVRWVlJYWNigteZsrwvg6+uLr6+vM2/voogK8aNnZBt255ayPquA6/pFu7okERERt+BUS42Pjw+JiYmkpqbaj1mtVlJTUxk2bFij8+Pj49m2bRsZGRn22w033MDo0aPJyMggNjaWqqoqqqqqMJsblmKxWOwBKDExEW9v7wavm5mZycGDBx2+bks3orvRBaVxNSIiIk3H6cX3UlJSmDp1KoMHD2bo0KEsXLiQsrIypk2bBsCUKVPo0KED8+fPx8/Pjz59+jS4vq6lpe64j48Pl19+OQ8//DD+/v507tyZ1atX89prr7FgwQIAQkJCmD59OikpKYSFhREcHMz999/PsGHDuPTSSy/k/bvEyJ7hLF23n7XaMkFERKTJOB1qbrnlFvLy8nj88cfJyclhwIABrFy50j54+ODBg41aXX7KsmXLmDNnDpMnT+bEiRN07tyZJ598kl/+8pf2c/72t79hNpuZNGkSFRUVjB07lhdeeMHZ8luEpC5h+FjMHCk8xf78Mrq2b+PqkkRERFo9k81ms7m6iIuhuLiYkJAQioqKCA4OdnU53Pby16TvK2DejZcwZVicq8sRERFpkZz5/G7W2U9ydiN6GLOg1uzW1G4REZGmoFDjIqNq16v5el8BVTXaMkFERORCKdS4yCUxwbQN8Ka0opqMQ4WuLkdERKTVU6hxEbPZxGX2LRPUBSUiInKhFGpcaGTtuJo0rVcjIiJywRRqXKhuH6iMQ4UUnapycTUiIiKtm0KNC3UI9adr+0CsNkjPKnB1OSIiIq2aQo2LjbSPq1EXlIiIyIVQqHGxkbVdUGl7NVhYRETkQijUuNil3drhZTZxoKCcgwXlri5HRESk1VKocbE2vl4M6tQWgLV71QUlIiJyvhRqWoC6LRPWassEERGR86ZQ0wLUrVezPiufam2ZICIicl4UalqAfh1DCfbzovh0NVuPFLm6HBERkVZJoaYFsJyxZUKatkwQERE5Lwo1LYR9XI3WqxERETkvCjUtxKja9Wq+O1hIaUW1i6sRERFpfRRqWojYsAA6twug2mrja22ZICIi4jSFmhZkpLqgREREzptCTQsyorvRBbVWWyaIiIg4TaGmBRnWrR1mE+zLK+NI4SlXlyMiItKqKNS0ICH+3gyIDQUgTV1QIiIiTlGoaWFG1M6CWqP1akRERJyiUNPCjKodLLxubz41VpuLqxEREWk9FGpamP6xobTx9aKwvIrtR7VlgoiIyLlSqGlhvC1mhnVrB8BadUGJiIicM4WaFkjr1YiIiDhPoaYFGlk7WHjTgZOUV2rLBBERkXOhUNMCxbULoEOoP1U1NjbsO+HqckRERFoFhZoWyGQyMapnXReUxtWIiIicC4WaFqpuy4S0vRpXIyIici4Ualqoy7q3w2SC3bml5BSddnU5IiIiLZ5CTQsVGuBDvw4hAKRpg0sREZGfpFDTgtXNgtLUbhERkZ+mUNOCjThjywSrtkwQERH5UQo1LdigTm0J8LGQX1rJzpxiV5cjIiLSoinUtGA+XmYu7WpsmZCmqd0iIiI/SqGmhRvRXevViIiInAuFmhaubhG+jdknOF1V4+JqREREWi6FmhauW/s2RAX7UVltZeN+bZkgIiJyNucVap5//nni4uLw8/MjKSmJjRs3ntN1y5Ytw2QyMWHChAbHTSaTw9tf/vIX+zlxcXGNHn/qqafOp/xWxWQy2Xft1no1IiIiZ+d0qFm+fDkpKSk88cQTbN68mf79+zN27FiOHz/+o9dlZ2fz0EMPMXLkyEaPHTt2rMFt6dKlmEwmJk2a1OC8efPmNTjv/vvvd7b8Vqluavea3VqvRkRE5GycDjULFixgxowZTJs2jd69e7No0SICAgJYunTpWa+pqalh8uTJzJ07l65duzZ6PCoqqsHtP//5D6NHj250blBQUIPzAgMDnS2/VaobLLwrp4S8kgoXVyMiItIyORVqKisr2bRpE8nJyfVPYDaTnJxMenr6Wa+bN28eERERTJ8+/SdfIzc3lxUrVjg896mnnqJdu3YMHDiQv/zlL1RXV5/1eSoqKiguLm5wa63atfHlkphgwFiIT0RERBrzcubk/Px8ampqiIyMbHA8MjKSXbt2ObwmLS2NJUuWkJGRcU6v8eqrrxIUFMTEiRMbHH/ggQcYNGgQYWFhrF+/njlz5nDs2DEWLFjg8Hnmz5/P3Llzz+k1W4ORPdqz/Wgxa/bkMWFgB1eXIyIi0uI4FWqcVVJSwh133MHixYsJDw8/p2uWLl3K5MmT8fPza3A8JSXF/nW/fv3w8fHhnnvuYf78+fj6+jZ6njlz5jS4pri4mNjY2PN8J643skc4i1ZnkbYnH5vNhslkcnVJIiIiLYpToSY8PByLxUJubm6D47m5uURFRTU6Pysri+zsbMaPH28/ZrVajRf28iIzM5Nu3brZH1u7di2ZmZksX778J2tJSkqiurqa7OxsevXq1ehxX19fh2GntUrs3BY/bzPHSyrYnVtKr6ggV5ckIiLSojg1psbHx4fExERSU1Ptx6xWK6mpqQwbNqzR+fHx8Wzbto2MjAz77YYbbmD06NFkZGQ0ajlZsmQJiYmJ9O/f/ydrycjIwGw2ExER4cxbaLX8vC0M7WJsmaBdu0VERBpzuvspJSWFqVOnMnjwYIYOHcrChQspKytj2rRpAEyZMoUOHTowf/58/Pz86NOnT4PrQ0NDARodLy4u5u233+aZZ55p9Jrp6els2LCB0aNHExQURHp6OrNnz+b222+nbdu2zr6FVmtk93DW7M5j7Z587h7ZeBaZiIiIJ3M61Nxyyy3k5eXx+OOPk5OTw4ABA1i5cqV98PDBgwcxm51f02/ZsmXYbDZuu+22Ro/5+vqybNky/vCHP1BRUUGXLl2YPXt2gzEznmBkz3D4BDbsL6CiugZfL4urSxIREWkxTDabzebqIi6G4uJiQkJCKCoqIjg42NXlnBebzcbQP6WSV1LBW3cnMbz7uQ2+FhERaa2c+fzW3k+tiMlkYmTdrt1ar0ZERKQBhZpWpm7LBA0WFhERaUihppWp2zJh+9FiCkq1ZYKIiEgdhZpWJiLYj/ioIGw2WJdV4OpyREREWgyFmlZoZG0XVJq6oEREROwUalqhET3aA9i3TBARERGFmlZpaFwYPl5mjhadJiuvzNXliIiItAgKNa2Qv4+FIXHGSsrqghIRETEo1LRSI2u7oNbu0Xo1IiIioFDTatVN7f56XwGV1VYXVyMiIuJ6CjWtVO/oYNoF+lBWWcN3B0+6uhwRERGXU6hppcxmE5fVttakacsEERERhZrWrG7LhDUaVyMiIqJQ05rVLcK37XAhheWVLq5GRETEtRRqWrHoEH+6R7TBaoP12jJBREQ8nEJNKzfSvmu3uqBERMSzKdS0cvWhJk9bJoiIiEdTqGnlkrq0w9ti4vDJUxwoKHd1OSIiIi6jUNPKBfp6MaiTsWXCWk3tFhERD6ZQ4wZG9azdMmG39oESERHPpVDjBuq2TEjPKqC6RlsmiIiIZ1KocQN9OoQQGuBNSUU1Ww4XurocERERl1CocQMWs4nLumlqt4iIeDaFGjcxQuvViIiIh1OocRN142oyDhVSfLrKxdWIiIhcfAo1biI2LIAu4YHUWG2ka8sEERHxQAo1bqRudeE0dUGJiIgHUqhxI3VdUGv3aL0aERHxPAo1bmRYt3ZYzCayC8o5dEJbJoiIiGdRqHEjQX7eDIwNBTQLSkREPI9CjZsZ2cPYMiFtr7qgRETEsyjUuJm69WrW7S2gxmpzcTUiIiIXj0KNm+nfMYQgPy+KTlWx7UiRq8sRERG5aBRq3IyXxczwbu0ASNMsKBER8SAKNW6oblzNGg0WFhERD6JQ44bqFuH77uBJSiuqXVyNiIjIxaFQ44Y6twukU1gAVTU2NuzTlgkiIuIZFGrclHbtFhERT6NQ46ZGassEERHxMAo1bmp4t3DMJsjKK+No4SlXlyMiItLszivUPP/888TFxeHn50dSUhIbN248p+uWLVuGyWRiwoQJDY6bTCaHt7/85S/2c06cOMHkyZMJDg4mNDSU6dOnU1paej7le4SQAG/6dQwFtGu3iIh4BqdDzfLly0lJSeGJJ55g8+bN9O/fn7Fjx3L8+PEfvS47O5uHHnqIkSNHNnrs2LFjDW5Lly7FZDIxadIk+zmTJ09m+/btfP7553z88cesWbOGmTNnOlu+RxlVN65mr0KNiIi4P5PNZnNqLf2kpCSGDBnCc889B4DVaiU2Npb777+fRx991OE1NTU1jBo1irvuuou1a9dSWFjIBx98cNbXmDBhAiUlJaSmpgKwc+dOevfuzTfffMPgwYMBWLlyJddeey2HDx8mJibmJ+suLi4mJCSEoqIigoODnXnLrdbG/Sf4+UvphAX68O3/JGM2m1xdkoiIiFOc+fx2qqWmsrKSTZs2kZycXP8EZjPJycmkp6ef9bp58+YRERHB9OnTf/I1cnNzWbFiRYNz09PTCQ0NtQcagOTkZMxmMxs2bHD4PBUVFRQXFze4eZqBnUIJ9LFwoqySHcc87/2LiIhncSrU5OfnU1NTQ2RkZIPjkZGR5OTkOLwmLS2NJUuWsHjx4nN6jVdffZWgoCAmTpxoP5aTk0NERESD87y8vAgLCzvr686fP5+QkBD7LTY29pxe3514W8wMq90yQVO7RUTE3TXr7KeSkhLuuOMOFi9eTHh4+Dlds3TpUiZPnoyfn98FvfacOXMoKiqy3w4dOnRBz9da1W2ZoKndIiLi7rycOTk8PByLxUJubm6D47m5uURFRTU6Pysri+zsbMaPH28/ZrVajRf28iIzM5Nu3brZH1u7di2ZmZksX768wfNERUU1GohcXV3NiRMnHL4ugK+vL76+vs68PbdUtwjft9knOVVZg7+PxcUViYiINA+nWmp8fHxITEy0D+AFI6SkpqYybNiwRufHx8ezbds2MjIy7LcbbriB0aNHk5GR0ahLaMmSJSQmJtK/f/8Gx4cNG0ZhYSGbNm2yH/vyyy+xWq0kJSU58xY8TtfwQDqE+lNZY2XDfm2ZICIi7suplhqAlJQUpk6dyuDBgxk6dCgLFy6krKyMadOmATBlyhQ6dOjA/Pnz8fPzo0+fPg2uDw0NBWh0vLi4mLfffptnnnmm0WsmJCQwbtw4ZsyYwaJFi6iqqmLWrFnceuut5zTzyZOZTCZGdA9n+beHSNuTzxW9In76IhERkVbI6VBzyy23kJeXx+OPP05OTg4DBgxg5cqV9sHDBw8exGx2fqjOsmXLsNls3HbbbQ4ff/PNN5k1axZjxozBbDYzadIk/v73vzv9Op5oZE8j1GiwsIiIuDOn16lprTxxnZo6J8sqGfS/n2OzwcbHxhARfGGDsEVERC6WZlunRlqntoE+9IkJATS1W0RE3JdCjYcYWTsLKk1bJoiIiJtSqPEQdVO71+7Jx0N6HEVExMMo1HiIxM5t8fe2kF9awa6cEleXIyIi0uQUajyEr5eFpK5hgFYXFhER96RQ40Hqt0zQuBoREXE/CjUepG6w8Mb9JzhdVePiakRERJqWQo0H6RHRhshgXyqqrXybfdLV5YiIiDQphRoPYmyZUNsFtVfjakRExL0o1HiYUT1rp3bv1rgaERFxLwo1Huay7kao2XGsmPzSChdXIyIi0nQUajxMeBtfekcbe2es0+rCIiLiRhRqPNDIM1YXFhERcRcKNR6ofsuEPG2ZICIibkOhxgMNiQvD18tMbnEFe4+XurocERGRJqFQ44H8vC0M7WJsmbBGXVAiIuImFGo8VN24mjTtAyUiIm5CocZD1S3C9/W+E1RUa8sEERFp/RRqPFR8VBDhbXw5VVXD5gOFri5HRETkginUeCiz2cSI7u0ASNOWCSIi4gYUajzYyB61+0BpsLCIiLgBhRoPVrdezbYjRZwsq3RxNSIiIhdGocaDRQb70SsyCJsN1mWptUZERFo3hRoPN8I+tVuhRkREWjeFGg935j5Q2jJBRERaM4UaD5fUpR0+FjNHCk+xP7/M1eWIiIicN4UaD+fvY2FwXFtAs6BERKR1U6iRM3btVqgREZHWS6FGGGnfMqGAqhqri6sRERE5Pwo1wiUxwbQN8Ka0opqMQ4WuLkdEROS8KNQIZrOJy7rXdkHt1pYJIiLSOinUCACjardM+O/2XE5VatduERFpfRRqBIAr4tsT4GMhM7eEyf/4WtsmiIhIq6NQIwBEBPnx6l1DCfbzYvPBQn62aD2HT5a7uiwREZFzplAjdkPiwnjn3uFEh/iRlVfGpBfXsyun2NVliYiInBOFGmmgZ2QQ7/1qOD0j25BbXMHNi9L5el+Bq8sSERH5SQo10kh0iD9v3zOcIXFtKTldzZSlG/l02zFXlyUiIvKjFGrEoZAAb16fnsTYSyKprLbyq7c283p6tqvLEhEROSuFGjkrP28LL0xO5BdJnbDZ4Pf/2c5f/5up3bxFRKRFUqiRH2Uxm3hyQh9SruoJwHNf7eWRd7dSre0URESkhVGokZ9kMpl4YEwPnprYF7MJ/v3tYWa+vkmL9ImISItyXqHm+eefJy4uDj8/P5KSkti4ceM5Xbds2TJMJhMTJkxo9NjOnTu54YYbCAkJITAwkCFDhnDw4EH741dccQUmk6nB7Ze//OX5lC/n6dahnXjpjsH4epn5ctdxfvGPrzmhRfpERKSFcDrULF++nJSUFJ544gk2b95M//79GTt2LMePH//R67Kzs3nooYcYOXJko8eysrIYMWIE8fHxrFq1iq1bt/L73/8ePz+/BufNmDGDY8eO2W9//vOfnS1fLtBVvSN5a0YSIf7efFe7SN+hE1qkT0REXM9kc3LUZ1JSEkOGDOG5554DwGq1Ehsby/3338+jjz7q8JqamhpGjRrFXXfdxdq1ayksLOSDDz6wP37rrbfi7e3N66+/ftbXveKKKxgwYAALFy50ply74uJiQkJCKCoqIjg4+LyeQ+rtPV7ClCUbOVp0moggX169aygJ0fp7FRGRpuXM57dTLTWVlZVs2rSJ5OTk+icwm0lOTiY9Pf2s182bN4+IiAimT5/e6DGr1cqKFSvo2bMnY8eOJSIigqSkpAahp86bb75JeHg4ffr0Yc6cOZSXn72FoKKiguLi4gY3aTrdI4J471eX0SsyiOMlFfx8UTrpWVqkT0REXMepUJOfn09NTQ2RkZENjkdGRpKTk+PwmrS0NJYsWcLixYsdPn78+HFKS0t56qmnGDduHJ999hk33XQTEydOZPXq1fbzfvGLX/DGG2/w1VdfMWfOHF5//XVuv/32s9Y6f/58QkJC7LfY2Fhn3qqcg6gQP/79y2EM7RJGSUU1U5duZMVWLdInIiKu4dWcT15SUsIdd9zB4sWLCQ8Pd3iO1WpMDb7xxhuZPXs2AAMGDGD9+vUsWrSIyy+/HICZM2far+nbty/R0dGMGTOGrKwsunXr1uh558yZQ0pKiv1+cXGxgk0zCPH35rW7hvLgsgxWbs9h1r82k196CVOHx7m6NBER8TBOhZrw8HAsFgu5ubkNjufm5hIVFdXo/KysLLKzsxk/frz9WF2I8fLyIjMzk9jYWLy8vOjdu3eDaxMSEkhLSztrLUlJSQDs3bvXYajx9fXF19f33N+cnDc/bwvPTx7EHz7czutfH+CJD7dzvOQ0D13dC5PJ5OryRETEQzjV/eTj40NiYiKpqan2Y1arldTUVIYNG9bo/Pj4eLZt20ZGRob9dsMNNzB69GgyMjKIjY3Fx8eHIUOGkJmZ2eDa3bt307lz57PWkpGRAUB0dLQzb0GaicVsYt6Nl/DQ1cYifc9/lcXD72ylSov0iYjIReJ091NKSgpTp05l8ODBDB06lIULF1JWVsa0adMAmDJlCh06dGD+/Pn4+fnRp0+fBteHhoYCNDj+8MMPc8sttzBq1ChGjx7NypUr+eijj1i1ahVgtPi89dZbXHvttbRr146tW7cye/ZsRo0aRb9+/c7zrUtTM5lMzLqyB+2DfHns/e95Z9NhCkoreH7yIAJ8mrWnU0RExPlQc8stt5CXl8fjjz9OTk4OAwYMYOXKlfbBwwcPHsRsdm75m5tuuolFixYxf/58HnjgAXr16sW7777LiBEjAKOF6IsvvrAHqNjYWCZNmsTvfvc7Z8uXi+CWIZ0Ib+PLfW9t5qvMPG5bvIFX7hxCWKCPq0sTERE35vQ6Na2V1qm5+DYdOMn0V7+hsLyKruGBvHrXUGLDAlxdloiItCLNtk6NiDMSO7flnV8Op0OoP/vyy5j44np2HNV6QSIi0jwUaqRZdY9ow3u/Gk58VBB5JRXc8lI667PyXV2WiIi4IYWappCXCZ7Ri3deIoP9WH7PMJJqF+m7c+k3fLz1qKvLEhERN6NQc6EKsuDFy+D1m+D4LldX02KF+Hvz6l1DubZvFJU1Vu7/13e8sm6/q8sSERE3olBzoY5sApMJ9n0FLw6HTx+FU4WurqpF8vO28Oxtg5gyrDM2G8z9aAdPr9yFh4xVFxGRZqZQc6H6/Rzu2wC9rgNbDWx4EZ4dBJv+CdYaV1fX4ljMJubecAkPj+0FwIursvjN21u0SJ+IiFwwhZqmENYVbnsLbn8PwntCeQF89Gt4+Qo4+LWrq2txTCYT943uzp9/1g+L2cR7m49w96vfUlZR7erSRESkFVOoaUrdx8C962HsfPANhpytsHQsvHs3FGtg7A/9fHAsi6ck4udtZvXuPH6x+GsKSitcXZaItAbVFcYvjdWVrq5EWhCFmqZm8YZhv4L7N8OgKYAJtr0Nzw6GNX+FqtOurrBFuTI+kn/NuJS2Ad5sOVzEzxalc7Cg3NVliUhLlrsdXh5t/NL4/BDY9g5Y1YUtWlG4+R39Dj59BA5tMO63jYOrn4T464wBxgJAVl4pU5Zs5EjhKcLb+PLPaUPo0yHE1WWJSEtitcKGRfDFH6DmB6260QPgqnnQ9XJXVCbNSCsKtyQxA+Gu/8LExRAUDSezYflkTQH/gW7tjUX6EqKDyS+t4NaXv2bdXi3SJyK1io/BGxPhv3OMQNPjanggA0b/D/i0gWMZ8NoN8MbPjJYc8UhqqbmYKkph7TOQ/hzUVILJAkn3wOWPgH+oa2pqYYpPVzHztW/5et8JvC0mnvn5AG7oH+PqskTElXZ8CB89AKdOgpcfXP2/MOTu+tbu0jxY82f4dilYqwET9L8NrvwfCOno0tLlwjnz+a1Q4won9sF//wcyPzHuB4TDmN/DwDvAbHFtbS1ARXUNKcu3sGLbMQB+f31vpo/o4uKqROSiqyiFlY/Ad28Y96P6waR/QPtejs8vyILUebDjA+O+xRcu/SWMmA3+bS9KydL0FGocaFGhps7eVFj5KOTvNu5H94dr/gydLnVtXS2A1Wpj3sc7+Of6bADuGdWVR8bFYzZrHJKIRzj8rTFz9OR+wASX/droavLyOYdrN8Hnj8OBNOO+XyiMegiGzABvv+asWpqBQo0DLTLUANRUwcbFsGo+VNTuYN33ZmPAW7Bnd7vYbDZeXJ3Fn1dmAjBxYAee/lk/vC0aCibitmqqjW761U8bC5oGd4SbFkGXkc49j80Gez6Dz5+AvJ3GsZBOcOXvjJ+xZv0caS0UahxosaGmTmkefDkPNr8O2MA7EEamwLBZHv+bxdvfHuLR97ZRY7Uxqmd7Xpw8iEBfr8Ynlh6HHf+BnR+BdwAkTjUGE6pLT6R1OLEf3psJhzca9/tMguueubCuI2sNZLwFX/0JSmrXC4vqC8lzjbXFpMVTqHGgxYeaOke/g09+W/+fum0cjP0T9LrWo6eAf5V5nF+9sZlTVTX06xjC0juHEN7GF8pPwM4P4fv3IHst2H6wVkVoZ2NA4cDbISDMNcWLyI+z2Yzg8elvobLUWLz0umeMbWiaSmW5sY1N2sL6VvGuVxit4tH9m+51pMkp1DjQakINGP/Bt71t9AmXGINl6Toarnn67APkPEDGoULu+uc3VJUV8ovgrTwY/T3+h9bUznao1SERLpkIpbmw+TU4XWgc9/KHfjfD0JnGb2ki0jKUn4CPHzRaWQE6DTe6m9p2bp7XKyuAtX81uv2tVcaxvj83uqWa6zXlgijUONCqQk2dH04BN3sZH8qeOAW8ohR2r6Rs87/x3p+KD2cEmai+RjP1JTcZLVt1Ksvh+3dgw8uQu63+eKfhMHQGJIw3VoAWEdfI+go+uNf45c3sBaMfg8sevDhdxif2w5f/a/yMALD4GAOJRz2kVt0WRqHGgVYZaup46hTwqlOw53P4/l3Y/V+oPmV/6IC5E+9UDCXVchmP3XEDI3qEn/15bDZjRecNLxldVXUtO0HRMPguSLwT2kQ073sRkXrVFcbU6/TnjPvtuhsLlHYYdPFrOfqdMZh4/2rjvm8IjJwNSb8Eb/+LX480olDjQKsONXU8YQp4dSVkfWkEmcxPjP71OmFdja6lPpMoCenBPa9vYn1WAd4WE3+9uT83Dujw089ffAw2vQLfvgJlx41jZm+jlWfoTOg42KPHLok0u9wd8N4MyP3euD/4LmMxPZ9A19Vks0FWqhFu6uoK7mC0HPW/zb1/eWwFFGoccItQA2eZAv5zuGpu650CXlNt/Ja0/T1j5tLpovrHQmKNwNFnkhHizggcFdU1/ObfW/h4qzHu6HfXJXD3yK7n9prVlUYf/saX6wdlg7GtxdCZRnjy8FlnIk3KaoWNLxnBoabCaHG+8TnodY2rK6tnrYGt/za6pYoPG8ciehszpXpcpV94XEShxgG3CTV1SvMgdW7tSputcAq4tQYOrDdaZHZ+COUF9Y+1iaoPMj/RcmK12vjjih28si4bgBkjuzDnmgTnFuk7+p0RFLe9U79JXkA7GDTV+C0yNPY83qCI2JXkwAe/MlpDwFhq4YbnICjStXWdTdVp4xeetX+t/yUrbqTxy2OHRNfW5oEUahxwu1BTpzVNAbda4fA3RpDZ8YExQ6lOQDj0vhH6TIROw5xq7rXZbLy0Zh9PfWpsEDphQAx//ll/fLycXFyrrAA2vwrfLKn/Lc1kNnZUHzrT+KHW0v5ORVq6nR/Dh/fDqROO921qycpPQNoCY7JB3S88l0w0xjSGnWOrsFwwhRoH3DbUQMueAm6zGcFr+3vw/fv1YQHALwQSbjCCTNwosDhYUM8J7246zCPvbqXaamNQp1AeTO7JyB7hmJz94VlTDbtXGk3l+9fUH2+fYMya6ncL+La5oFpF3F5FqbGj9ubXjPtRfWHiPyAi3rV1nY/Cg8bifVuWATZjHN7gu+Dy30Lgj0xSkCahUOOAW4eaOi1lCrjNBrnba4PMe7V7t9TyCTJaPvpMNILXuezj4oRVmcf51ZubKa+sASAhOpiZo7pwfb+Y89te4fguoxl6yzKoKjOO+YbAwMnGb5vtujVh9SJu4vC3xmDgE/sw9m16AEb/rsn/v190OduMMUF13Wg+QTDi13Dpr1w70NnNKdQ44BGhpo7DKeCPG6vqNuco/rzd9UEmP7P+uHcA9BxnBJnuVzX7mJ/DJ8tZkraf5d8csoebmBA/7hrRhVuHdqKNoy0WfsrpImPF042L4URW/fHuVxnBsXuy9pIRqak2umtWPVW7b1OH2n2bRrm6sqa1b5XRMn5si3G/TRSMngMDbr/gFmdpTKHGAY8KNXX2fgEr5zTvFPAT++u7ls5c4M7ia8wW6DPRCDQu+C2msLySNzcc5JV12eSXGv3hQX5eTE7qzLTL4ogMPo9wZbXCvi+NPvY9nwG1/33adjG6pgZM9ryFEd2Z1WpM8d2/xtiGozTXWFo/YTzEDGod40IulhP74f17jDWhwBh7cv2CC9u3qSWzWo3xgV/OM7qnAMJ7QfIfjBld+rfRZBRqHPDIUAO1U8BfNn5zaqop4EWHYfv7RovM0c31x81e0O1KY9ZSr2vBr2X8PZ+uquGD747w8tp97MszupC8LSZuHNCBmaO60jMy6Pye+MQ+Y1Dxd6/Xz5DwDjDG3AydAZGXNNE7kIvGZoO8TCPA7F8N2Wlw6qTjc4M7QPz1kHC9sUq1p/6GbrPBln8ZExYqS4x9m679q7Fvkyd8sFdXGD8H1vzFGAwNxmSHq/4IsUNcW5ubUKhxwGNDTZ0LnQJekmvMWPr+PTj0df1xk9loWr5kovHbawteXtxqtfHlruO8vGYfG7NP2I+P7tWemaO6cWnXMOcHFQNUlhkDtTe8DMe31x+PG2mEm17Xee4HXktnsxnhNHut0Rqzf239oox1fNoYH1JdRhkrT2d+aqx0XTfGCsA/zAjyCeONlpzWsKxCUyg/AR/PNn42gPH3dNNLnrmH0ukiY7PMr1+A6tPGsYTxMOYPEN7dlZW1ego1Dnh8qKnjcAr4fMfNpWUFsPM/RpA5sO6MHbBN0Hm4sZZM7xtb5RYD3x08yctr9rFyew51/wP6dQxh5qiujLskCq/zGVRssxlr72x8yZjGajPG8xDcwZgpMWgqtGnfdG9Czk/hoYYh5swZeWBMO45NMkJMl1HGgow/3COs6pQxrmLnR8bYtTNbc3zaGF2v8dcb67G0kBbLJrdvNbz/Syg5arTSXjEHRszW6rtFR2DVn4wxeDYrmCzGVixXPNoqf1a2BAo1DijUnKFuCvhnv4fSHONYtyth3FPQJhJ2rTD6ivetqv9gBug4xGiRuWRC6129+Aey88v4R9o+3v72MBXVRmiLDfNn+mVd+PmQWAJ8zrOFpegIfLsUNv0TyvONYxYfo2tu6Awt4HUxleTWdyftX9twNh4Y03M7DoEuI40Q03EIePme+/PXVBuhf9fHRpgtOVr/mMWnfgxOr2vdY/qvw32bXta/6R/K3WG0ju9eadz3DoTh9xs3LQnhFIUaBxRqHKgoqZ0C/nz9FHCT2fi6TlS/M3bAdt8m5YLSCl5LP8Br6dmcLK8CIDTAmzsu7cyUYXG0D3LiQ+5M1RWw/QOj9ebIpvrjHRJh6D1GQHTmA1R+WvmJhi0xZ87EA+PfeMyg+hATm9R0A9mtVqM1dNdHRitOwd6Gr9tpmBFw4q9vnStVH98J786onxSQeKex2KemM59ddprxC2Td+MPACLjiEaPl9octgOKQQo0DCjU/oiALPvtd/RTw9gn1QcbD+oJPVdbwzubD/GPtPg4UlAPg42Vm0qCO3D2yC93aX8BvWIc3GYO2t79XHxwD2xsfDInTIOQcNuSUxk4XGd1+dSHmzFl4AJiMhd/qupM6Dbs4XUJ1g453fmSEnLrpv3WiBxiDjBNucP0imT/FZjP+7X7+uDFeJKCdsc1B/LWurqx1sNmMcUep82rX7sFo4RrzhBFyPWFA9QVQqHFAoeYc5O4w+sNb+g/Yi6DGauOz7Tm8tGYfGYcKAePnTnJCJPeM6srguAsYEF2aZ2zH8O1SKD5iHDNZjA+4ofcY45X0Q+7sKsvgYLoRYPavgWMZZ4z3qtU+ob4lpvNlLWMA+8kDRtfuro+NEMYZP3rb9TA+3BKub3lTxX+4b1P3ZLjxhZa7b1NLVl1pdEmvfrq+W7rjUGM2amySxiOdhUKNAwo1cj5sNhvfZBuDir/YWb9X1aBOocwc1Y2rekdicWbzzDPVVEPmCmPW1IG0+uORfYxxN31vVrM+GJsLHt5YH2KObAJrVcNzwrrVh5i4kS1/QGZpntEyuvMjY+zame8nuKOx6nbCeKNVyZUz53atMPZtKi8wBlBf9Ufj32ZLCl2t0eliWP+sMS6pqrz+uJe/Md7GJ9AYcO4TeMYt6Iyvax9zeG6bhvfdICgp1DigUCMXau/xUv6xdh/vbT5CZY3RMtAlPJDpI7rws8SO+HlfwA+P3O3GasVbl9f/kPMLgZ7XGDOm/Nsa04b92xqtDmfe9/Z3rw+Z6kpj/MH+2sG9hzbWbyZYJyS2vjspbmTr7ro7XWRMEd/5UcuZKl5ZZizcuflV435kX5jUSvdtaslKcmDVfGOpDWt187yGl/8ZAeiHQenMYz94zNdBiKr78yIHJYUaBxRqpKkcLznNq+uzeePrgxSdMn7Dbhfow5RhcUwZ1pm2gRewv82pk/XbMfxwls7ZWHx/EHRCHYefHx5rKWupWGuMLqS6lpiDXzf8YAdjVl5dgOkyyliKwJ2CXB37VPGPjVY8V0wVP7ypdt+mLMBkzNa58nca0N6cqiuMvfsqS4xAWVkGlaXGnxWl9V83eKz0LOfW3j9z5mpTqwtKdUHH94xAFD0ARj3UpC+nUOOAQo00tbKKapZ/c4glafs5UngKAD9vMz8fHMvdI7rSqV3A+T+51QpZXxof9qdO1t/KT5xx/8SF/Xbn5f+DoPPD4OOodajthX+4Wa3GIoV1IebAeqgoaniOf5jRnRQ3ErpcDuE93DPE/Jiaaji43mjBuRhTxa01sHaB0XLgzvs2eQKbzQhKjQLQGUGoosRBUHL05xnB6lyCUver4PZ3mvTtKNQ4oFAjzaW6xson3+fw8posvj9ibEVhNsE1faKZOaor/WNDm+eFbTbjB80Pg86pk1B+0sGxM867kN/ivANrg07bHw8/Zx47dbJ+/6T9a+uXk6/jGwJxl9W3xET01gahZ/rJqeLDjUHG5ztV/GQ2vHdP/Wrhl9wE1//NffdtEuc1CkpnaTEKijbGhDWhZg81zz//PH/5y1/Iycmhf//+PPvsswwdOvQnr1u2bBm33XYbN954Ix988EGDx3bu3MkjjzzC6tWrqa6upnfv3rz77rt06tQJgNOnT/Ob3/yGZcuWUVFRwdixY3nhhReIjDy3EfgKNdLcbDYb6VkFvLRmH6t359mPD+0Sxj2jujK6VwTm8x1U3JRsNmMfsEYtPz9sDTrR8NjpwsazjM6XdyB0HlYfYqL7u8WAxouibqp4XcC5kKniNhtsWQafPGx0ffgEwXV/NfYv87SWMWmxmjXULF++nClTprBo0SKSkpJYuHAhb7/9NpmZmUREnH3GQXZ2NiNGjKBr166EhYU1CDVZWVkMHTqU6dOnc9tttxEcHMz27du59NJL7c957733smLFCv75z38SEhLCrFmzMJvNrFu37pzqVqiRi2lXTjEvr9nHhxlHqbYa/8W6R7Rh5siu3DgwBl+vVvgBbrUa3UT2oHPScfj54bFThUZ3SackiKsd3NthkBYeayrnO1X81Elj36bt7xv3Yy+FiS8Z45VEWpBmDTVJSUkMGTKE554zlsi2Wq3ExsZy//338+ijjzq8pqamhlGjRnHXXXexdu1aCgsLG4SaW2+9FW9vb15//XWH1xcVFdG+fXveeustfvaznwGwa9cuEhISSE9P59JLL/3JuhVqxBWOFZ3in+uyeWvDQUoqjPEvEUG+3HlZHJOHdiYkwAM+2K01RouANvVsfnVTxXd9bAw4PnN18DOnilur4T/3GeskmSz1+zbpeyQtkDOf3051WldWVrJp0yaSk5Prn8BsJjk5mfT09LNeN2/ePCIiIpg+fXqjx6xWKytWrKBnz56MHTuWiIgIkpKSGoSeTZs2UVVV1eB14+Pj6dSp01lft6KiguLi4gY3kYstOsSfOdcmsG7OlTx2bTxRwX4cL6ngzyszGf5UKvM+2sHhk+U//UStmdmiD8uLpU17SJwKk9+Gh7Ng0hLoPcHo7is+bGzX8er18PoEI9CEdYPpn8PlD+t7JG7BqVCTn59PTU1No3EskZGR5OTkOLwmLS2NJUuWsHjxYoePHz9+nNLSUp566inGjRvHZ599xk033cTEiRNZvXo1ADk5Ofj4+BAaGnrOrzt//nxCQkLst9jYVrjPiriNYD9vZo7qxprfjuaZm/sTHxVEWWUNS9ft5/K/rOLXy77j+yNFP/1EIufKLxj6/gx+/ir8dh/cthwG3F4/+HfQVLhnDXTURpTiPpo1mpeUlHDHHXewePFiwsMdTzm0Wo2BhzfeeCOzZ88GYMCAAaxfv55FixZx+eWXn9drz5kzh5SUFPv94uJiBRtxOR8vM5MSOzJxUAdW785j8dp9rNtbwH8yjvKfjKNc1r0dM0d1Y1SPcEwaqClNxdsPeo0zbjX/B9WnjMXVRNyMU6EmPDwci8VCbm5ug+O5ublERUU1Oj8rK4vs7GzGjx9vP1YXYry8vMjMzCQ2NhYvLy969+7d4NqEhATS0oyl46OioqisrKSwsLBBa83ZXhfA19cXX18tFiUtk8lk4opeEVzRK4LvjxTx8pp9rNh2jHV7C1i3t4D4qCBmjurK+P4xeFs0tVmakMULLAo04p6c+mnp4+NDYmIiqamp9mNWq5XU1FSGDRvW6Pz4+Hi2bdtGRkaG/XbDDTcwevRoMjIyiI2NxcfHhyFDhpCZmdng2t27d9O5c2cAEhMT8fb2bvC6mZmZHDx40OHrirQmfTqE8PfbBrLqoSuYdlkcAT4WduWUkPLvLQya9zk/X5TO4//5nn9tPMh3B09SXtlMy6mLiLRyTnc/paSkMHXqVAYPHszQoUNZuHAhZWVlTJs2DYApU6bQoUMH5s+fj5+fH3369GlwfV1Ly5nHH374YW655RZGjRrF6NGjWblyJR999BGrVq0CICQkhOnTp5OSkkJYWBjBwcHcf//9DBs27JxmPom0BrFhATwx/hIeHNOTNzYc4J/rs8krqWBj9gk2ZtcvVmcyQeewABKig4mPCiY+OoiEqGA6tvVvGevgiIi4iNOh5pZbbiEvL4/HH3+cnJwcBgwYwMqVK+2Dhw8ePIjZyZVAb7rpJhYtWsT8+fN54IEH6NWrF++++y4jRoywn/O3v/0Ns9nMpEmTGiy+J+JuQgK8uW90d2aO6sre46Xsyilm57ESdh4rZldOCXklFWQXlJNdUM6n39cPlG/j60WvqCDio4KIjw6md3QQPSODCPLzgGnjIiJomwSRVie/tIJdx0rsYWdXTjF7ckvtO4f/UGyYP/FRwSREB5NQG3g6hwWoVUdEWgXt/eSAQo24s6oaK/vzy+ytOTuPFbPrWAk5xacdnu/vbaFXVBAJ0UFGF1Zt2AnxV6uOiLQsCjUOKNSIJzpZVlkfcnKMwJOZU0JFteNWnQ6h/rUBJ8g+ZieuXQBemoElIi6iUOOAQo2IocZqY39+mRFyzujGOlJ4yuH5vl5mekae0apTOzC5baDPRa5cRDyRQo0DCjUiP67oVBWZOSUNBiZn5pRwqqrG4fmRwb721py6wNO1faDW1RGRJqVQ44BCjYjzrFYbB0+UsyunmB3HSthVO2bn4AnH+1X5WMx0j2hjb81JiDZadsLbaCFMETk/CjUOKNSINJ3Simoyz5h9ZXRjlVBa4XhhwM7tAvj1mB5MGNBBs65ExCkKNQ4o1Ig0L6vVxpHCUw1nYOWUkF1QRt1PmT4dgnnsmgSGd3e8F5yIyA8p1DigUCPiGqUV1byWns2LX2VRUtuSc2V8BHOuiadHpPYgEpEfp1DjgEKNiGsVlFbw99Q9vLnhINVWG2YT3Dq0Ew8m9yAiyM/V5YlIC6VQ44BCjUjLsC+vlKdX7uK/23MBCPCx8MvLu3H3yC4E+Di9c4uIuDmFGgcUakRalo37T/DkJzvZcqgQMKaI/+aqXkxK7IhFg4lFpJZCjQMKNSItj81m4+Otx3h65S4OnzQW/4uPCuKxaxMY1bO9i6sTkZZAocYBhRqRlquiuobX1h/g2S/3UHzaGEw8skc4j12bQEK0/r+KeDKFGgcUakRavsLySp79ci+vpWdTVWPDZIKbEzuSclUvokI0mFjEEynUOKBQI9J6HCgo48//zWTF1mMA+HmbmTmyKzMv70YbXw0mFvEkCjUOKNSItD6bDpzkT5/sZNOBkwCEt/Fl9lU9uGVwrHYOF/EQCjUOKNSItE42m43/bs/hqU93kV1g7DnVPaINj10bz+heEZhMmikl4s4UahxQqBFp3Sqrrby54QB/T93DyfIqAIZ1bcf/XJdAnw4hLq5ORJqLQo0DCjUi7qHoVBUvrNrLK+uyqay2AjBxYAd+M7YXHUL9XVydiDQ1hRoHFGpE3Mvhk+X89b+ZfJBxFAAfLzPTR3Th3iu6Eezn7eLqRKSpKNQ4oFAj4p62Hi7kyRU72bD/BABhgT78ekwPfpHUCW8NJhZp9RRqHFCoEXFfNpuN1J3Hmf/pTrLyygDoEh7Io9fEc3XvSA0mFmnFFGocUKgRcX/VNVaWfXOIv32+m4KySgCGxLXlsWsTGNiprYurE5HzoVDjgEKNiOcoOV3FS6v38Y+0fZyuMgYTj+8fw2/H9iI2LMDF1YmIMxRqHFCoEfE8x4pO8cxnu3l382FsNvCxmJk6vDOzRvcgJECDiUVaA4UaBxRqRDzXjqPF/OmTnaTtzQcgxN+b+6/szh3DOuPrZXFxdSLyYxRqHFCoEfFsNpuN1bvzmP/JLjJzSwDoFBbAI+PiubZvlAYTi7RQCjUOKNSICECN1cY7mw7xzGe7OV5SAcDATqH8z7UJDI4Lc3F1IvJDCjUOKNSIyJnKK6tZvGY/L63JoryyBoBxl0TxyDXxdAkPdHF1IlJHocYBhRoRceR48Wn+9sVuln9zCKsNvMwmbr+0Mw+M6UFYoI+ryxPxeAo1DijUiMiP2Z1bwvxPdvJVZh4AQX5e3De6O3cOj8PPW4OJRVxFocYBhRoRORfr9ubz5Iqd7DhWDECHUH8eHtuLG/rHYDZrMLHIxaZQ44BCjYicK6vVxvvfHeGvn2VyrOg0AH07hDDn2niGdW2nmVIiF5FCjQMKNSLirNNVNSxJ28+Lq7IoragGoEdEG36W2JGbBnYgItjPxRWKuD+FGgcUakTkfOWXVvB/X+zh398eoqLa2HbBbILLe7bnZ4mxjEmI0LgbkWaiUOOAQo2IXKji01Ws2HqMdzYdZtOBk/bjIf7e3NA/hpsHd6RvhxB1T4k0IYUaBxRqRKQp7csr5d3Nh3l30xFyik/bj/eMNLqnJgzsQESQuqdELpRCjQMKNSLSHGqsNtbtzeedTYf57/Yce/eUxWyq7Z7qyJiECO0xJXKeFGocUKgRkeZWdKque+oQmw8W2o+HBnhzY/8YfpYYS58OweqeEnGCQo0DCjUicjFl5ZXy7qbDvLe5YfdUr8ggfpbYkRsHxqh7SuQcOPP5bT6fF3j++eeJi4vDz8+PpKQkNm7ceE7XLVu2DJPJxIQJExocv/POOzGZTA1u48aNa3BOXFxco3Oeeuqp8ylfRKTZdWvfht+Oi2fdo1fy6l1DGd8/Bh8vM5m5JTz5yU6Gzf+S6f/8hk+3HaOytstKRC6Ml7MXLF++nJSUFBYtWkRSUhILFy5k7NixZGZmEhERcdbrsrOzeeihhxg5cqTDx8eNG8crr7xiv+/r69vonHnz5jFjxgz7/aCgIGfLFxG5qOrG1lzesz1Fp6r4eOtR3tl0mO8OFpK66zipu47TNsCbGwd04GeJHbkkRt1TIufL6VCzYMECZsyYwbRp0wBYtGgRK1asYOnSpTz66KMOr6mpqWHy5MnMnTuXtWvXUlhY2OgcX19foqKifvS1g4KCfvIcEZGWKsTfm8lJnZmc1Jm9x43ZU+9tPkxucQX/XJ/NP9dnEx9V2z01oAPtgxr/ciciZ+dU91NlZSWbNm0iOTm5/gnMZpKTk0lPTz/rdfPmzSMiIoLp06ef9ZxVq1YRERFBr169uPfeeykoKGh0zlNPPUW7du0YOHAgf/nLX6iurj7r81VUVFBcXNzgJiLSUnSPaMMj4+JZ/+gY/jltCNf3i8bHy8yunBL+d8VOhs1P5e5Xv2Xl9znqnhI5R0611OTn51NTU0NkZGSD45GRkezatcvhNWlpaSxZsoSMjIyzPu+4ceOYOHEiXbp0ISsri8cee4xrrrmG9PR0LBZjGuQDDzzAoEGDCAsLY/369cyZM4djx46xYMECh885f/585s6d68zbExG56CxmE1f0iuCKXhEUlVfxUW33VMahQr7YmcsXO3PVPSVyjpya/XT06FE6dOjA+vXrGTZsmP34b3/7W1avXs2GDRsanF9SUkK/fv144YUXuOaaawBjUHBhYSEffPDBWV9n3759dOvWjS+++IIxY8Y4PGfp0qXcc889lJaWOhx/U1FRQUVFhf1+cXExsbGxmv0kIq3C3uMlvL3pMO9vPsLxkvqfZXXdUxMGdiC8jbqnxP05M/vJqZaa8PBwLBYLubm5DY7n5uY6HOuSlZVFdnY248ePtx+zWo1mVC8vLzIzM+nWrVuj67p27Up4eDh79+49a6hJSkqiurqa7OxsevXq1ehxX19fh2FHRKQ16B4RxJxrEnj46l6srV3c7/Ptufbuqac+3cXo+Ah+ltiR0b0i8PE6r8msIm7FqVDj4+NDYmIiqamp9mnZVquV1NRUZs2a1ej8+Ph4tm3b1uDY7373O0pKSvi///s/YmNjHb7O4cOHKSgoIDo6+qy1ZGRkYDabf3TGlYhIa+dlMTO6VwSje0VQWF7JR7V7T205VMjnO3L5fEcuYYE+3DggprZ7KsTVJYu4jNOzn1JSUpg6dSqDBw9m6NChLFy4kLKyMvtsqClTptChQwfmz5+Pn58fffr0aXB9aGgogP14aWkpc+fOZdKkSURFRZGVlcVvf/tbunfvztixYwFIT09nw4YNjB49mqCgINLT05k9eza33347bdu2vZD3LyLSaoQG+HDHpZ2549LO7M4tMRb3++4IeSUVvLIum1fWZdM7Orh29lQM7dQ9JR7G6VBzyy23kJeXx+OPP05OTg4DBgxg5cqV9sHDBw8exGw+92ZQi8XC1q1befXVVyksLCQmJoarr76aP/7xj/buI19fX5YtW8Yf/vAHKioq6NKlC7NnzyYlJcXZ8kVE3ELPyCDmXJvAw2N7sXZPbffUjlx2HCtm3sc7+NMnO7myrnsqPgJvi7qnxP1pmwQRETdRWF7JR1uO8vamw2w9XGQ/3i7QhxsHdODmwR1JiNbPP2ldtPeTAwo1IuJJMnNKahf3O0J+af3sqUtigpk0qCPX94/W3lPSKijUOKBQIyKeqLrGypo9ebyz6TBf7DhOZY0xA9Vsgku7tuP6fjGM6xNFWKCPiysVcUyhxgGFGhHxdCfLKvlo61He23yEjEOF9uNeZhOXdQ/n+n7RXH1JFCH+3q4rUuQHFGocUKgREal36EQ5H289xsdbj7L9aP02Mj4WM6N6tmd8/2iSEyIJ9HV6PolIk1KocUChRkTEsX15pXy89RgfbTnKnuOl9uN+3mbGxEdyfb9oRsdH4OdtcWGV4qkUahxQqBER+WmZOSV8vPUoH205SnZBuf14oI+Fq3pHMr5/DCN7tNcKxnLRKNQ4oFAjInLubDYb248W89GWo3y89RhHCk/ZHwv282Jcnyiu7xfD8G7t8NIaONKMFGocUKgRETk/NpuN7w4V8tGWo6zYeqzBBpthgT5cUxtwhnYJw2LWDuLStBRqHFCoERG5cDVWG99kn+DjrUf5ZFsOJ8oq7Y9FBPlybd9oxvePYVCnUEwmBRy5cAo1DijUiIg0reoaK+n7Cvhoy1FWfp9D8elq+2MdQv25vl801/eLoU+HYAUcOW8KNQ4o1IiINJ/Kaitr9+Tx8dZjfLY9h7LKGvtjce0CuL5fDOP7x9ArKsiFVUprpFDjgEKNiMjFcbqqhlWZx/loyzFSd+Vyuspqf6xHRJvagBNN1/ZtXFiltBYKNQ4o1IiIXHxlFdV8sTOXj7ceY3Vmnn2bBoDe0cGM7x/D9f2iiQ0LcGGV0pIp1DigUCMi4lpFp6r4fEcuH289StqefKqt9R8/A2JD7WNwokK00abUU6hxQKFGRKTlOFlWycrtOXy05Shf7yugLt+YTDCkcxjj+0dzTd9owtv4urZQcTmFGgcUakREWqbjJaf5dFsOH289yjfZJ+3HzSYY3s3YaHNcnyhCA7STuCdSqHFAoUZEpOU7WniKT7YZ+1BtOVxkP+5lNjGyRzjj+8dwVe9Igvy0k7inUKhxQKFGRKR1OVhQzsfbjvLRlmPsPHbGTuJeZq7o2Z5r+0bTuV0AbQN8aBvoQ7Cfl9bDcUMKNQ4o1IiItF57j5faN9rMyitzeI7FbCLU35u2gT60DfAmNMCHsAAfQgO9aVv3dUDd4/XnaGuHlk2hxgGFGhGR1s9ms7GrdifxtD355JdWcrK8kvIzFvtzVoi/d30ICqwNPj/42mgNMr4ODfDG18vShO9KfoxCjQMKNSIi7quiuobC8ipOlBkhp+7rwvJKTpZXcbL2+MnyKuPPssoG2zo4K9DHctYQZPxptATVdY21DfDG39ui7rHz4Mznt9dFqklERKTZ+HpZiAy2EBl87mvcVNdYKTxVRWF5JSfKqmrDkPG1EYYafn2y3PjaaoOyyhrKKk9xpPCUEzWa7S09YbVdYHVfd2zrT6ewQOLCA4gM8sOsLrHzolAjIiIeyctiJryNr1Nr4VitNopPVzVo8akLOyd+8HVh3TnllVTV2KiotpJTfJqc4tM/+hq+XmY6hQXQuV0AndsF1v8ZFkCHtv54W8wX+tbdlkKNiIjIOTKbTYQGGN1LXQg8p2tsNhtllTUNusDODEEFpRUcOnmKAwVlHD55iopqK3uOl7LneGmj57KYTXQI9a8NOgHEtQusDUDGn/4+nj3WR6FGRESkGZlMJtr4etHG1+sn97iqqrFytPAUBwrKOVBQxoGCcrILyjl4wvi6otrKwRPlHDxRzto9ja+PDPa1t+rEhRtBJ65dIJ3aBRDi7/5r+yjUiIiItBDeFnNtl1Mg0L7BY1arjeMlFfawc+BEmRF4CsrJLiij5HQ1ucUV5BZXsHH/iUbPHRrgXR942gXQqV1g7Z8BtG/j6xaDmDX7SUREpJWz2WwUlldx4MSZLTxltYGnnPzSih+9PsDHYm/V6VwbdOq6tmJC/V26lo9mP4mIiHgQk8lkTB0P9GFAbGijx8sqqjlwRjdWXZdWdn45x4pOUV5Zw66cEnbllDS61ttiIrbtDwcuG193bOvfotbsUagRERFxc4G+XvSOCaZ3TOOWjorqGg6fPMXB2nE8RuAxWnoOnzhFZY2Vffll7MsvA/IaXGsyQUxI/cDl/h1DuXVop4v0rhpTqBEREfFgvl4WurVvQ7f2bRo9VmO1cazolL0b68CJ+i6tgwVllFXWcKTQWK9nfVYBh0+eUqgRERGRlsdiNtGxbQAd2wYwvHvDx2w2G/mllfZurAMnyunY1t81hdZSqBERERGnmUwm2gf50j7Il8TOYa4uBwAtSygiIiJuQaFGRERE3IJCjYiIiLgFhRoRERFxCwo1IiIi4hYUakRERMQtKNSIiIiIW1CoEREREbdwXqHm+eefJy4uDj8/P5KSkti4ceM5Xbds2TJMJhMTJkxocPzOO+/EZDI1uI0bN67BOSdOnGDy5MkEBwcTGhrK9OnTKS0tPZ/yRURExA05HWqWL19OSkoKTzzxBJs3b6Z///6MHTuW48eP/+h12dnZPPTQQ4wcOdLh4+PGjePYsWP227/+9a8Gj0+ePJnt27fz+eef8/HHH7NmzRpmzpzpbPkiIiLippwONQsWLGDGjBlMmzaN3r17s2jRIgICAli6dOlZr6mpqWHy5MnMnTuXrl27OjzH19eXqKgo+61t27b2x3bu3MnKlSv5xz/+QVJSEiNGjODZZ59l2bJlHD161Nm3ICIiIm7IqVBTWVnJpk2bSE5Orn8Cs5nk5GTS09PPet28efOIiIhg+vTpZz1n1apVRERE0KtXL+69914KCgrsj6WnpxMaGsrgwYPtx5KTkzGbzWzYsMHh81VUVFBcXNzgJiIiIu7LqVCTn59PTU0NkZGRDY5HRkaSk5Pj8Jq0tDSWLFnC4sWLz/q848aN47XXXiM1NZWnn36a1atXc80111BTUwNATk4OERERDa7x8vIiLCzsrK87f/58QkJC7LfY2Fhn3qqIiIi0Ms26S3dJSQl33HEHixcvJjw8/Kzn3Xrrrfav+/btS79+/ejWrRurVq1izJgx5/Xac+bMISUlxX6/qKiITp06qcVGRESkFan73LbZbD95rlOhJjw8HIvFQm5uboPjubm5REVFNTo/KyuL7Oxsxo8fbz9mtVqNF/byIjMzk27dujW6rmvXroSHh7N3717GjBlDVFRUo4HI1dXVnDhxwuHrgjFGx9fX136/7i9FLTYiIiKtT0lJCSEhIT96jlOhxsfHh8TERFJTU+3Tsq1WK6mpqcyaNavR+fHx8Wzbtq3Bsd/97neUlJTwf//3f2cNGIcPH6agoIDo6GgAhg0bRmFhIZs2bSIxMRGAL7/8EqvVSlJS0jnVHhMTw6FDhwgKCsJkMp3rWz4nxcXFxMbGcujQIYKDg5v0ucV5+n60LPp+tCz6frQ8+p78OJvNRklJCTExMT95rtPdTykpKUydOpXBgwczdOhQFi5cSFlZGdOmTQNgypQpdOjQgfnz5+Pn50efPn0aXB8aGgpgP15aWsrcuXOZNGkSUVFRZGVl8dvf/pbu3bszduxYABISEhg3bhwzZsxg0aJFVFVVMWvWLG699dZzepNgDGju2LGjs2/XKcHBwfoH2YLo+9Gy6PvRsuj70fLoe3J2P9VCU8fpUHPLLbeQl5fH448/Tk5ODgMGDGDlypX2wcMHDx7EbD738ccWi4WtW7fy6quvUlhYSExMDFdffTV//OMfG3Qfvfnmm8yaNYsxY8ZgNpuZNGkSf//7350tX0RERNyUyXYuI2/kRxUXFxMSEkJRUZFSdgug70fLou9Hy6LvR8uj70nT0d5PTcDX15cnnniiQcuSuI6+Hy2Lvh8ti74fLY++J01HLTUiIiLiFtRSIyIiIm5BoUZERETcgkKNiIiIuAWFGhEREXELCjUX6PnnnycuLg4/Pz+SkpLYuHGjq0vyWPPnz2fIkCEEBQURERHBhAkTyMzMdHVZAjz11FOYTCYefPBBV5fi0Y4cOcLtt99Ou3bt8Pf3p2/fvnz77beuLssj1dTU8Pvf/54uXbrg7+9Pt27d+OMf/3hO+xvJ2SnUXIDly5eTkpLCE088webNm+nfvz9jx45ttE+VXByrV6/mvvvu4+uvv+bzzz+nqqqKq6++mrKyMleX5tG++eYbXnrpJfr16+fqUjzayZMnueyyy/D29ubTTz9lx44dPPPMM7Rt29bVpXmkp59+mhdffJHnnnuOnTt38vTTT/PnP/+ZZ5991tWltWqa0n0BkpKSGDJkCM899xxg7IMVGxvL/fffz6OPPuri6iQvL4+IiAhWr17NqFGjXF2ORyotLWXQoEG88MIL/O///i8DBgxg4cKFri7LIz366KOsW7eOtWvXuroUAa6//noiIyNZsmSJ/dikSZPw9/fnjTfecGFlrZtaas5TZWUlmzZtIjk52X7MbDaTnJxMenq6CyuTOkVFRQCEhYW5uBLPdd9993Hdddc1+H8irvHhhx8yePBgbr75ZiIiIhg4cCCLFy92dVkea/jw4aSmprJ7924AtmzZQlpaGtdcc42LK2vdnN77SQz5+fnU1NTY97yqExkZya5du1xUldSxWq08+OCDXHbZZY02VZWLY9myZWzevJlvvvnG1aUIsG/fPl588UVSUlJ47LHH+Oabb3jggQfw8fFh6tSpri7P4zz66KMUFxcTHx+PxWKhpqaGJ598ksmTJ7u6tFZNoUbc0n333cf3339PWlqaq0vxSIcOHeLXv/41n3/+OX5+fq4uRzCC/uDBg/nTn/4EwMCBA/n+++9ZtGiRQo0L/Pvf/+bNN9/krbfe4pJLLiEjI4MHH3yQmJgYfT8ugELNeQoPD8disZCbm9vgeG5uLlFRUS6qSgBmzZrFxx9/zJo1a+jYsaOry/FImzZt4vjx4wwaNMh+rKamhjVr1vDcc89RUVGBxWJxYYWeJzo6mt69ezc4lpCQwLvvvuuiijzbww8/zKOPPsqtt94KQN++fTlw4ADz589XqLkAGlNznnx8fEhMTCQ1NdV+zGq1kpqayrBhw1xYmeey2WzMmjWL999/ny+//JIuXbq4uiSPNWbMGLZt20ZGRob9NnjwYCZPnkxGRoYCjQtcdtlljZY42L17N507d3ZRRZ6tvLwcs7nhR7DFYsFqtbqoIveglpoLkJKSwtSpUxk8eDBDhw5l4cKFlJWVMW3aNFeX5pHuu+8+3nrrLf7zn/8QFBRETk4OACEhIfj7+7u4Os8SFBTUaCxTYGAg7dq10xgnF5k9ezbDhw/nT3/6Ez//+c/ZuHEjL7/8Mi+//LKrS/NI48eP58knn6RTp05ccsklfPfddyxYsIC77rrL1aW1bja5IM8++6ytU6dONh8fH9vQoUNtX3/9tatL8liAw9srr7zi6tLEZrNdfvnltl//+teuLsOjffTRR7Y+ffrYfH19bfHx8baXX37Z1SV5rOLiYtuvf/1rW6dOnWx+fn62rl272v7nf/7HVlFR4erSWjWtUyMiIiJuQWNqRERExC0o1IiIiIhbUKgRERERt6BQIyIiIm5BoUZERETcgkKNiIiIuAWFGhEREXELCjUiIiLiFhRqRERExC0o1IiIiIhbUKgRERERt6BQIyIiIm7h/wFAlTzAbnWG6gAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "this_history_df[['loss', 'val_loss']].plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "2342a31f-4891-4834-851a-b90b8b729c3c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Axes: >"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "this_history_df[['auc', 'val_auc']].plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "cee80637-cabf-4974-b476-08c5fba165ff",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Axes: >"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "this_history_df[['auc_roc', 'val_auc_roc']].plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "e184b0ca-1d31-4798-88a0-4e590b67062e",
   "metadata": {},
   "outputs": [],
   "source": [
    "this_model.save(Path(setup.temp_path).joinpath('25_DATA_model.keras'),save_format='tf')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d326c9ea-4759-4418-a64d-56c151049c25",
   "metadata": {},
   "source": [
    "## Predictions on all data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "cd6910ee-5ffe-4def-9aa8-7dba76446b08",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  46/9167 [..............................] - ETA: 20s "
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-05-20 13:45:48.039289: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:114] Plugin optimizer for device_type GPU is enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "9167/9167 [==============================] - 21s 2ms/step\n"
     ]
    }
   ],
   "source": [
    "all_predictions = this_model.predict([X[features_numeric]]+ [X[f] for f in features_naics])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "98a2cce1-b7d3-4ebc-8869-80705297e145",
   "metadata": {},
   "outputs": [],
   "source": [
    "all_predictions_df = pd.DataFrame(all_predictions, index=X.index) \\\n",
    "    .set_axis(['predict_prob'], axis=1) \\\n",
    "    .reset_index() \\\n",
    "    .merge(sba_loans[['target', 'LoanNr_ChkDgt', 'dset', 'dset_naics_holdout', 'NAICS']], \n",
    "           on='LoanNr_ChkDgt')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "3fc7da9c-6bd2-4d1b-b77f-b62679825d5d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>predict_prob</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>predict_prob</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.302722</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>target</th>\n",
       "      <td>0.302722</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              predict_prob    target\n",
       "predict_prob      1.000000  0.302722\n",
       "target            0.302722  1.000000"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "all_predictions_df[['predict_prob', 'target']].corr(method='spearman')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e79afc4e-cb67-4064-9b19-dc52500baef2",
   "metadata": {},
   "source": [
    "##### Threshold Tune & Binary Predictions\n",
    "Using training probability predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "ccd0bc23-fbc0-40d0-8d8b-f2302405a875",
   "metadata": {},
   "outputs": [],
   "source": [
    "all_pred_train = all_predictions_df[all_predictions_df['dset'] == 'train']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "df36d9e9-f35e-46b3-a5be-08081f998319",
   "metadata": {},
   "outputs": [],
   "source": [
    "thresh_tune_data = sbnn_metrics.get_f1_frame(all_pred_train['target'], \n",
    "                                        all_pred_train['predict_prob'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "de634156-caf6-4a52-b5d7-44b8e986aa1d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>thresh</th>\n",
       "      <th>f1</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0.22</td>\n",
       "      <td>0.457660</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>0.24</td>\n",
       "      <td>0.457179</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.20</td>\n",
       "      <td>0.456342</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    thresh        f1\n",
       "11    0.22  0.457660\n",
       "12    0.24  0.457179\n",
       "10    0.20  0.456342"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "thresh_tune_data.sort_values('f1', ascending=False, inplace=True)\n",
    "thresh_tune_data.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "c746b880-d781-4bc1-9df2-89211127c22a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.22"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_thresh = thresh_tune_data['thresh'].iloc[0]\n",
    "best_thresh"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3c0ac5e2-be0d-42f0-960d-1f81b4a5d4e2",
   "metadata": {},
   "source": [
    "##### Append binary predictions to probability predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "641da72e-1bca-4fe7-96c4-6c029b521bd7",
   "metadata": {},
   "outputs": [],
   "source": [
    "all_predictions_df['predict_bin'] = sbnn_metrics.get_binary_predictions(all_predictions_df['predict_prob'], best_thresh)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "9f6ecdf4-ac01-4c39-8315-e8583daa337e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "predict_bin\n",
       "0    0.562795\n",
       "1    0.437205\n",
       "Name: proportion, dtype: float64"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "all_predictions_df['predict_bin'].value_counts(normalize=True, dropna=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "cbfed830-7ff8-40f6-a691-6b894b26d6bc",
   "metadata": {},
   "outputs": [],
   "source": [
    "all_predictions_df.to_parquet(Path(setup.temp_path).joinpath('25_DATA_predictions.parquet'))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "eb0ecbd2-abff-46d8-8c70-f7bccd363b69",
   "metadata": {},
   "source": [
    "## Metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "be4ceefe-c7a8-45bd-8573-2daefe99a1c2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>dset</th>\n",
       "      <th>accuracy_score</th>\n",
       "      <th>f1_score</th>\n",
       "      <th>precision_score</th>\n",
       "      <th>recall_score</th>\n",
       "      <th>average_precision_score</th>\n",
       "      <th>roc_auc_score</th>\n",
       "      <th>dset_naics_holdout</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>test</td>\n",
       "      <td>0.645170</td>\n",
       "      <td>0.441055</td>\n",
       "      <td>0.323236</td>\n",
       "      <td>0.694027</td>\n",
       "      <td>0.354774</td>\n",
       "      <td>0.714913</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>train</td>\n",
       "      <td>0.650994</td>\n",
       "      <td>0.457660</td>\n",
       "      <td>0.334248</td>\n",
       "      <td>0.725549</td>\n",
       "      <td>0.390071</td>\n",
       "      <td>0.738473</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>val</td>\n",
       "      <td>0.637766</td>\n",
       "      <td>0.439292</td>\n",
       "      <td>0.320891</td>\n",
       "      <td>0.696156</td>\n",
       "      <td>0.357649</td>\n",
       "      <td>0.711592</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>test</td>\n",
       "      <td>0.638310</td>\n",
       "      <td>0.436973</td>\n",
       "      <td>0.319554</td>\n",
       "      <td>0.690806</td>\n",
       "      <td>0.352795</td>\n",
       "      <td>0.708043</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>test</td>\n",
       "      <td>0.656572</td>\n",
       "      <td>0.448060</td>\n",
       "      <td>0.329591</td>\n",
       "      <td>0.699485</td>\n",
       "      <td>0.365265</td>\n",
       "      <td>0.727972</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    dset  accuracy_score  f1_score  precision_score  recall_score  \\\n",
       "0   test        0.645170  0.441055         0.323236      0.694027   \n",
       "1  train        0.650994  0.457660         0.334248      0.725549   \n",
       "2    val        0.637766  0.439292         0.320891      0.696156   \n",
       "0   test        0.638310  0.436973         0.319554      0.690806   \n",
       "1   test        0.656572  0.448060         0.329591      0.699485   \n",
       "\n",
       "   average_precision_score  roc_auc_score  dset_naics_holdout  \n",
       "0                 0.354774       0.714913                 NaN  \n",
       "1                 0.390071       0.738473                 NaN  \n",
       "2                 0.357649       0.711592                 NaN  \n",
       "0                 0.352795       0.708043                 0.0  \n",
       "1                 0.365265       0.727972                 1.0  "
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "metrics_dset_df = all_predictions_df.groupby('dset') \\\n",
    "    .apply(lambda x: sbnn_metrics.dset_metrics(x.target, x.predict_bin, x.predict_prob)) \\\n",
    "    .reset_index()\n",
    "metrics_test_df = all_predictions_df[all_predictions_df['dset'] == 'test'] \\\n",
    "    .groupby(['dset', 'dset_naics_holdout']) \\\n",
    "    .apply(lambda x: sbnn_metrics.dset_metrics(x.target, x.predict_bin, x.predict_prob)) \\\n",
    "    .reset_index()\n",
    "metrics_df = pd.concat([metrics_dset_df, metrics_test_df])\n",
    "metrics_df.to_csv(Path(setup.temp_path).joinpath('25_REPORT_metrics.csv'), index=True)\n",
    "metrics_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "52e382b5-e9dc-47a1-9bf4-6d0b6b4a0c61",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tf_p39",
   "language": "python",
   "name": "tf_p39"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.18"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}