Update app.py

app.py

@@ -1,18 +1,30 @@
 import gradio as gr
 import joblib
+import numpy as np
+from sklearn.feature_extraction.text import CountVectorizer
+from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, roc_auc_score
+from sklearn.model_selection import train_test_split
+from sklearn.svm import SVC
+from sklearn.neighbors import KNeighborsClassifier
+from sklearn.linear_model import LogisticRegression
+from sklearn.ensemble import RandomForestClassifier
 
-# Load models (Ensure the correct paths to where your models are stored)
+# Load your pre-trained models (make sure the models are uploaded to the Hugging Face Space)
 logreg_model = joblib.load('best_lr_model.pkl')
 knn_model = joblib.load('best_knn_model.pkl')
 svc_model = joblib.load('best_svc_model.pkl')
 rf_model = joblib.load('best_rf_model.pkl')
 
+# Load the vectorizer (assuming the same vectorizer used during training)
+vectorizer = joblib.load('vectorizer.pkl')  # You must upload this file as well
+
 # Prediction function
 def predict_sentiment(review_text, model_name):
-    # Reshape input to match expected 2D array format (single sample, single feature)
-    review_text_reshaped = [review_text]  # This ensures the input is in a 2D array (1 sample, N features)
-    
-    # Select the appropriate model
+    # Preprocess the review text with the same vectorizer
+    review_features = vectorizer.transform([review_text])  # Transform review to features
+    review_features = review_features.toarray()  # Ensure it's in array form
+
+    # Select the model based on the input
     if model_name == "Logistic Regression":
         model = logreg_model
     elif model_name == "K-Nearest Neighbors":
@@ -21,48 +33,40 @@ def predict_sentiment(review_text, model_name):
         model = svc_model
     elif model_name == "Random Forest":
         model = rf_model
-    else:
-        return "Model not found", {}
-
-    # Assuming the models support `predict_proba` for probabilities
-    try:
-        prob = model.predict_proba(review_text_reshaped)  # Adjust according to your preprocessing steps
-        positive_prob = prob[0][1]
-        negative_prob = prob[0][0]
-        
-        # Predict the sentiment class (based on the highest probability)
-        predicted_class = "Positive Feedback" if positive_prob > negative_prob else "Negative Feedback"
-        
-        return predicted_class, {"Positive Comment": f"{positive_prob * 100:.2f}%", "Negative Comment": f"{negative_prob * 100:.2f}%"}
-
-    except Exception as e:
-        return f"Error: {e}", {}
+    
+    # Use the model to predict sentiment probabilities
+    prob = model.predict_proba(review_features)  # Use predict_proba to get probabilities
+    
+    # Return positive and negative sentiment probabilities
+    return {"Positive Comment": prob[0][1], "Negative Comment": prob[0][0]}
 
-# Gradio interface setup
+# Gradio Interface setup
 def create_interface():
+    # Create a dropdown for model selection
     model_dropdown = gr.Dropdown(
         choices=["Logistic Regression", "K-Nearest Neighbors", "Support Vector Machine", "Random Forest"],
         label="Select Model"
     )
 
-    # Textbox for review input
-    review_input = gr.Textbox(lines=2, placeholder="Enter your review comment here...")
+    # Create a textbox for the review input
+    review_input = gr.Textbox(
+        lines=2, placeholder="Enter your review comment here...", label="Review Comment"
+    )
 
-    # Gradio output for predicted sentiment and probabilities
-    sentiment_output = gr.Textbox(label="Predicted Sentiment Class")
-    prob_output = gr.JSON(label="Predicted Probability")
+    # Create a JSON output to display predicted sentiment and probabilities
+    output = gr.JSON()
 
-    # Create Gradio interface
+    # Create the Gradio interface
     interface = gr.Interface(
         fn=predict_sentiment, 
         inputs=[review_input, model_dropdown], 
-        outputs=[sentiment_output, prob_output],
+        outputs=output, 
         live=True,  # Optionally update in real-time
         description="Sentiment Analysis Model",
-        allow_flagging="never"  # Disable flagging option if you prefer
     )
 
     return interface
 
-# Launch the interface
-create_interface().launch(share=True)  # Add share=True to create a public link
+# Launch the Gradio interface
+if __name__ == "__main__":
+    create_interface().launch()