Add confidence interval calculation to application

app/price_predict.py

@@ -3,18 +3,21 @@
 from fastapi.templating import Jinja2Templates
 import pandas as pd
 import joblib
+import numpy as np
+from scipy.stats import norm
 
 # Load the trained model and scaler
 model = joblib.load("../model_components/final_catboost_model.joblib")
 scaler = joblib.load("../model_components/scaler.joblib")  
 X_columns = joblib.load("../model_components/X_columns.joblib")  
+y = joblib.load("../model_components/y.joblib")
+y_pred = joblib.load("../model_components/y_pred.joblib")
 numerical_columns = ["Area", "Rooms"]  
 
 # Initialize FastAPI and templates
 app = FastAPI()
 templates = Jinja2Templates(directory="templates")
 
-# Define the prediction function
 def predict_house_price(model, area, elevator, rooms, floor, region, year_of_building):
     if floor == 0:
         floor_category = "parter"
@@ -58,12 +61,18 @@ def predict_house_price(model, area, elevator, rooms, floor, region, year_of_bui
     predicted_price = model.predict(input_data)
     return predicted_price[0]
 
-# Route for rendering the HTML form
+def calculate_individual_confidence_interval(y_true, y_pred, prediction, alpha=0.05):
+    residuals = y_true - y_pred
+    standard_error = np.std(residuals, ddof=1)
+    margin_of_error = norm.ppf(1 - alpha / 2) * standard_error
+    lower_bound = prediction - margin_of_error
+    upper_bound = prediction + margin_of_error
+    return lower_bound, upper_bound
+
 @app.get("/", response_class=HTMLResponse)
 async def home(request: Request):
     return templates.TemplateResponse("index.html", {"request": request})
 
-# Route for handling the form submission and returning the prediction
 @app.post("/predict/", response_class=HTMLResponse)
 async def predict(
     request: Request,
@@ -72,18 +81,26 @@ async def predict(
     rooms: int = Form(...),
     floor: int = Form(...),
     region: str = Form(...),
-    year_of_building: int = Form(...)
+    year_of_building: int = Form(...),
+    alpha: float = Form(0.05),
 ):
     if elevator not in [0, 1]:
         raise HTTPException(status_code=400, detail="Invalid value for elevator. Must be 0 or 1.")
+
     predicted_price = predict_house_price(model, area, elevator, rooms, floor, region, year_of_building)
+    lower, upper = calculate_individual_confidence_interval(y, y_pred, predicted_price, alpha)
+    confidence_level = (1 - alpha) * 100
+
     return templates.TemplateResponse("index.html", {
         "request": request,
         "predicted_price": f"{predicted_price:,.2f}",
+        "confidence_interval": f"[{lower:,.2f}, {upper:,.2f}]",
+        "confidence_level": confidence_level,
         "area": area,
         "elevator": elevator,
         "rooms": rooms,
         "floor": floor,
         "region": region,
-        "year_of_building": year_of_building
+        "year_of_building": year_of_building,
+        "alpha": alpha,
     })

app/templates/index.html

@@ -15,8 +15,8 @@
             border-radius: 8px;
             box-shadow: 0 0 10px rgba(0, 0, 0, 0.1);
             width: 300px;
-            margin: 20px auto; /* Center the container */
-            overflow: hidden; /* Prevent horizontal scrolling */
+            margin: 20px auto;
+            overflow: hidden;
         }
         h1 {
             text-align: center;
@@ -61,15 +61,19 @@ <h1>House Price Prediction</h1>
         <form action="/predict/" method="post">
             <label for="area">Area (m²):</label>
             <input type="number" id="area" name="area" step="0.01" value="{{ area or '' }}" required>
+
             <label for="elevator">Elevator:</label>
             <select id="elevator" name="elevator" required>
                 <option value="1" {% if elevator == 1 %}selected{% endif %}>Yes</option>
                 <option value="0" {% if elevator == 0 %}selected{% endif %}>No</option>
             </select>
+
             <label for="rooms">Number of Rooms:</label>
             <input type="number" id="rooms" name="rooms" value="{{ rooms or '' }}" required>
+
             <label for="floor">Floor:</label>
             <input type="number" id="floor" name="floor" value="{{ floor or '' }}" required>
+
             <label for="region">Region:</label>
             <input type="text" id="region" name="region" list="regions" value="{{ region or '' }}" required>
             <datalist id="regions">
@@ -101,13 +105,21 @@ <h1>House Price Prediction</h1>
                 <option value="Za Cukrownią">
                 <option value="Zemborzyce">
             </datalist>
+
             <label for="year_of_building">Year of Building:</label>
             <input type="number" id="year_of_building" name="year_of_building" value="{{ year_of_building or '' }}" required>
+
+            <label for="alpha">Significance Level (α):</label>
+            <input type="number" id="alpha" name="alpha" step="0.01" min="0" max="1" value="{{ alpha or 0.05 }}" required>
+
             <button type="submit">Predict Price</button>
         </form>
+
         {% if predicted_price %}
             <div class="result">
                 <h2>Predicted Price: {{ predicted_price }} PLN</h2>
+                <p style="text-align: center;">{{ confidence_level }}% Confidence Interval</p>
+                <p style="text-align: center;">{{ confidence_interval }}</p>
             </div>
         {% endif %}
     </div>