Update app.py

app.py

@@ -2,10 +2,13 @@ import gradio as gr
 import numpy as np
 import joblib
 
-# Load saved objects
-model = joblib.load('model.pkl')
-scaler = joblib.load('scaler.pkl')
-label_enc = joblib.load('label_encoder.pkl')
+# Load models and encoder
+model_binary = joblib.load("model_binary_cv.pkl")
+model_multiclass = joblib.load("model_multiclass_cv.pkl")
+label_enc = joblib.load("label_encoder_multiclass.pkl")
+
+# Define scaler
+scaler = joblib.load("scaler.pkl")
 
 def predict_diabetes(
     Age,
@@ -23,16 +26,15 @@ def predict_diabetes(
     smoking_status,
     family_history_cardiovascular_disease
 ):
-    # Convert radio inputs to numeric
+    # Convert inputs
     alcohol_consumption = int(alcohol_consumption)
     family_history_diabetes = int(family_history_diabetes)
 
-    # One-hot encode categorical features
     Gender_FEMALE = 1 if Gender == "Female" else 0
     Gender_MALE = 1 if Gender == "Male" else 0
 
     dietary_non_vegetarian = 1 if dietary_habits == "Non-vegetarian" else 0
-    dietary_non_vegetarian_dup = 1 if dietary_habits == "Non-vegetarian" else 0  # for duplicate col with space
+    dietary_non_vegetarian_dup = 1 if dietary_habits == "Non-vegetarian" else 0  # duplicate column
     dietary_vegetarian = 1 if dietary_habits == "Vegetarian" else 0
 
     smoking_no = 1 if smoking_status == "No" else 0
@@ -41,11 +43,10 @@ def predict_diabetes(
     family_history_cvd_no = 1 if family_history_cardiovascular_disease == "No" else 0
     family_history_cvd_yes = 1 if family_history_cardiovascular_disease == "Yes" else 0
 
-    # Dummy patient ID and binary_status
-    patient_id = 0
-    binary_status = 0
+    patient_id = 0  # Dummy
+    # No binary_status in input — it's what we're predicting
 
-    # Match training input order (21 features)
+    # Construct input feature array for binary model (drop target fields)
     input_data = np.array([[
         patient_id,
         Age,
@@ -66,29 +67,32 @@ def predict_diabetes(
         smoking_no,
         smoking_yes,
         family_history_cvd_no,
-        family_history_cvd_yes,
-        binary_status
+        family_history_cvd_yes
     ]])
 
-    # Scale input
     input_scaled = scaler.transform(input_data)
 
-    # Predict and decode
-    pred_enc = model.predict(input_scaled)
-    pred_label = label_enc.inverse_transform(pred_enc)[0]
+    # Step 1: Binary prediction
+    risk_pred = model_binary.predict(input_scaled)[0]
+    if risk_pred == 0:
+        return "✅ Prediction: No Diabetes Risk (Normal)"
+
+    # Step 2: Multiclass prediction (only if at risk)
+    status_pred_enc = model_multiclass.predict(input_scaled)[0]
+    status_pred = label_enc.inverse_transform([status_pred_enc])[0]
 
-    return f"🔍 Predicted Status: {pred_label}"
+    return f"⚠️ At Risk: {status_pred}"
 
+# Gradio Interface
 with gr.Blocks() as demo:
-    gr.Markdown("# 🩺 Diabetes Predictor")
+    gr.Markdown("# 🩺 Diabetes Risk and Type Predictor")
     gr.Markdown(
         """
         Developed by Dr. Vinod Kumar Yata's research group  
         School of Allied and Healthcare Sciences, Malla Reddy University, Hyderabad, India
         
-        ⚠️ **Warning:**  
-        This is an experimental tool and should not be used for medical diagnosis.  
-        Always consult a licensed healthcare provider for medical advice.
+        ⚠️ **Note:** This tool predicts both diabetes **risk** and type.  
+        Not a substitute for medical advice.
         """
     )