Create app.py

app.py

@@ -0,0 +1,68 @@
+import joblib
+import numpy as np
+import pandas as pd
+import gradio as gr
+
+# --- 1. Load Model and Features ---
+# Ensure these two files are uploaded to your Hugging Face Space folder!
+try:
+    model = joblib.load('ckd_model.joblib')
+    # This list ensures input fields are created and data is fed in the correct order (24 features)
+    FEATURE_COLUMNS = joblib.load('model_features.joblib') 
+except FileNotFoundError:
+    raise FileNotFoundError("Model or feature files not found. Ensure 'ckd_model.joblib' and 'model_features.joblib' are uploaded.")
+
+# --- 2. Define the Prediction Function ---
+def predict_ckd(*inputs):
+    """
+    Takes 24 inputs (features) from the Gradio interface and returns the prediction.
+    """
+    # Convert inputs (which come as a tuple from Gradio) into a NumPy array
+    # and reshape it to match the model's required input format (1 sample, 24 features)
+    input_data = np.array(inputs).reshape(1, -1)
+    
+    # 1. Make the raw prediction (0 or 1)
+    prediction = model.predict(input_data)[0]
+    
+    # 2. Get the probability for the "Not CKD" class (class 1)
+    # This is often more informative than just a 0/1 result
+    probability_not_ckd = model.predict_proba(input_data)[0][1] 
+    
+    # Determine the final output text and confidence
+    if prediction == 0:
+        result_text = "Positive for Chronic Kidney Disease (CKD)"
+        confidence = 1 - probability_not_ckd # Confidence in CKD
+        color = "red"
+    else: # prediction == 1
+        result_text = "Negative for Chronic Kidney Disease (CKD)"
+        confidence = probability_not_ckd # Confidence in Not CKD
+        color = "green"
+
+    # Return prediction with formatted confidence
+    return f"<h2 style='color: {color};'>Prediction: {result_text}</h2><p>Confidence: {confidence:.2f} ({confidence*100:.0f}%)</p>"
+
+
+# --- 3. Create the Gradio Interface Components ---
+
+# Dynamically generate the 24 input components based on the feature list
+input_components = []
+for feature in FEATURE_COLUMNS:
+    # Use a generic Textbox for input, as Gradio will automatically handle numeric types
+    # You might customize these later (e.g., using gr.Slider for age, gr.Radio for binary inputs)
+    input_components.append(gr.Textbox(label=feature.upper(), placeholder=f"Enter value for {feature}"))
+
+# --- 4. Launch the Gradio Interface ---
+# The output component is set to HTML to allow for colored text output
+output_component = gr.HTML(label="Prediction Result")
+
+# Combine the function, inputs, and outputs into a Gradio interface
+iface = gr.Interface(
+    fn=predict_ckd,
+    inputs=input_components,
+    outputs=output_component,
+    title="CKD Prediction Model (Random Forest)",
+    description="Enter the 24 clinical parameters below to predict the risk of Chronic Kidney Disease. **Note: For binary features (yes/no, present/notpresent), use the encoded numerical values (0 or 1) used during training.**"
+)
+
+# Launch the app (Hugging Face Spaces runs this automatically)
+iface.launch(share=False)