sens_analysis_PE_method.py

from elicitation import parameter_elicitation_utilities_linear
import numpy as np
import matplotlib.pyplot as plt

import pysmile
import pysmile_license
import numpy as np
import pandas as pd

from df_plot import plot_df 
from info_value_to_net import info_value_to_net
from get_info_values import mutual_info_measures, cond_kl_divergence
from save_info_values import save_info_values
from plots import plot_cond_mut_info, plot_relative_cond_mut_info

from preprocessing import preprocessing

from network_functions import calculate_network_utilities, new_screening_strategy, old_screening_strategy, compare_strategies, create_folders_logger
from simulations import plot_classification_results
from plots import plot_estimations_w_error_bars, plot_screening_counts


import yaml
with open('config.yaml', 'r') as file:
    cfg = yaml.safe_load(file)

import logging
import datetime 
import os

np.seterr(divide='ignore', invalid = 'ignore')





def sens_analysis_PE_method(label = '', output_dir = 'logs', logger = None, log_dir = None):

    if logger == None:
        logger, log_dir = create_folders_logger(label=label, date = False, time = False, output_dir=output_dir)
    else:
        if not os.path.exists(log_dir):
            os.makedirs(log_dir)


    if cfg["new_test"] == True:
        file_location = "outputs/linear_rel_point_cond_mut_info_elicitFalse_newtestTrue/decision_models/DM_screening_rel_point_cond_mut_info_linear_new_test.xdsl"
    else:
        file_location = "outputs/linear_rel_point_cond_mut_info_elicitFalse_newtestFalse/decision_models/DM_screening_rel_point_cond_mut_info_linear.xdsl"

    net = pysmile.Network()
    net.read_file(file_location)


    PE_info_array = np.array(cfg["full_example"]["PE_info_array"])
    PE_cost_array = np.array(cfg["full_example"]["PE_cost_array"])

    logger.info(f"PE_info_array: {PE_info_array}")
    logger.info(f"PE_cost_array: {PE_cost_array}")

    net.update_beliefs()
    rho_comfort = net.get_node_value("Value_of_comfort")[2]

    df_test = pd.read_csv("private/df_2016.csv")
    df_test = preprocessing(df_test)
    df_test = df_test.rename(columns = {"Hyperchol.": "Hyperchol_"})

    # This function performs a sensitivity analysis on the PE method by varying its parameters.
    # The motivation is the depending on the relationship between PE cost and PE info, the
    # optimal screening strategy may change. We will explore how the distribution of recommended
    # screening strategies changes as we vary the parameters of the PE method.


    fig, axes = plt.subplots(len(PE_info_array), len(PE_cost_array), figsize=(16, 16))

    for i, param1 in enumerate(PE_info_array):
            for j, param2 in enumerate(PE_cost_array):
                # Call the custom function with the current combination of parameters
                params = parameter_elicitation_utilities_linear(net,PE = 0.7, PE_info = param1, PE_cost = param2, rho_comfort = rho_comfort, value_function = "rel_point_cond_mut_info", logging = None)
                logger.info(f"Params: {params}")
            
                net.set_mau_expressions(node_id = "U", expressions = [f"Max(0, Min({params[0]} - {params[1]}*Exp( - {params[2]} * V), 1))"])
                net.update_beliefs()

                _ , counts, possible_outcomes = calculate_network_utilities(net, df_test, full_calculation = True)

                #code from plots.py function plot_screening_counts()
                bars1 = axes[i,j].bar(possible_outcomes, counts, color = 'steelblue', alpha = 0.3, label = 'Number of tests')
                for bar in bars1:
                    axes[i,j].text(bar.get_x() + bar.get_width()/2, bar.get_height() + 3000, str(bar.get_height()), ha='center', color='black', fontsize=15)

                axes[i,j].legend()

                axes[i,j].set_ylim(0, 350000)
                axes[i,j].set_xticks(range(len(possible_outcomes)), possible_outcomes, rotation = 45)
                axes[i,j].set_xlabel("Screening outcome")
                axes[i,j].set_ylabel("Number of tests")
                axes[i,j].set_title("PE_info = " + str(param1) + " and PE_cost = " + str(param2))



                logger.info(f"PE_info: {param1}, PE_cost: {param2}, DONE!")

                plt.tight_layout()  
                if cfg["new_test"] == True:
                    plt.savefig(f"{log_dir}/sens_analysis_screening_counts_new_test.png")
                else:
                    plt.savefig(f"{log_dir}/sens_analysis_screening_counts.png")


    for handler in logger.handlers:
        handler.close()          # Close the handler
        logger.removeHandler(handler)  # Remove the handler from the logger

    plt.close(fig)


if __name__ == "__main__":
    sens_analysis_PE_method()
# Compare this snippet from sens_analysis_PE_method.py: