Update app.py

app.py

@@ -344,7 +344,7 @@ def predict_price(symbol):
         }
         tech_summary = "\n".join([f"{k}: {v}" for k, v in indicators.items()])
 
-        # Prepare LSTM features
+        # Prepare features for LSTM
         features = np.column_stack([
             df['Open'], df['High'], df['Low'], df['Close'],
             df['RSI'], df['RSI'], df['RSI']
@@ -366,7 +366,7 @@ def predict_price(symbol):
         last_close = float(df['Close'].iloc[-1])
         confidence = round(abs(predicted_price - last_close) * 0.05, 2)
 
-        # Adjust with YouTube Sentiment
+        # YouTube sentiment adjustment
         yt_sent = get_youtube_sentiment(symbol)
         yt_adj = predicted_price * (0.02 * yt_sent)
         adjusted_price = round(predicted_price + yt_adj, 2)
@@ -377,7 +377,7 @@ def predict_price(symbol):
         elif predicted_price < last_close * 0.98:
             reco = "Sell"
 
-        # Sentiment & Info
+        # Sentiment & Fundamentals
         industry = get_industry(symbol)
         stock_sent, stock_headlines = get_headlines_sentiment(symbol)
         industry_sent, industry_headlines = get_headlines_sentiment(industry)
@@ -391,7 +391,7 @@ def predict_price(symbol):
 📣 Sentiment score: {round(stock_sent, 2)} | Industry: {round(industry_sent, 2)}
 📋 Financials: {', '.join([f"{k}: {v}" for k, v in financials.items()])}"""
 
-        # Chart 1 – Prediction vs Actual
+        # Chart 1 – Actual vs Predicted
         plt.figure(figsize=(8, 4))
         plt.plot(df.index, df['Close'], label='Actual', marker='o')
         plt.plot(df.index, predicted_prices, label='Predicted', marker='x')
@@ -403,7 +403,7 @@ def predict_price(symbol):
         buf1.seek(0)
         chart1 = PIL.Image.open(buf1)
 
-        # Chart 2 – 1-min live
+        # Chart 2 – Intraday
         intraday = yf.download(symbol, period='1d', interval='1m')
         plt.figure(figsize=(8, 4))
         plt.plot(intraday.index, intraday['Close'], color='orange')
@@ -434,7 +434,7 @@ def predict_price(symbol):
         buf_sent.seek(0)
         sentiment_chart = PIL.Image.open(buf_sent)
 
-        # SHAP-style Feature Impact Chart
+        # 📊 Feature Influence Chart
         sample_input = scaled.reshape(1, 60, 7)
         base_pred = model.predict(sample_input)[0][0]
         impacts = []
@@ -444,9 +444,17 @@ def predict_price(symbol):
             new_pred = model.predict(perturbed)[0][0]
             impacts.append(new_pred - base_pred)
 
+        # Simplify RSI to 1 feature
+        feature_names = ['Open', 'High', 'Low', 'Close', 'RSI']
+        impact_values = [impacts[0], impacts[1], impacts[2], impacts[3], np.mean(impacts[4:])]
+        colors = ['green' if v >= 0 else 'red' for v in impact_values]
+
         plt.figure(figsize=(8, 4))
-        plt.barh(['Open', 'High', 'Low', 'Close', 'RSI1', 'RSI2', 'RSI3'], impacts)
-        plt.title("🧠 Feature Impact on LSTM Prediction")
+        bars = plt.barh(feature_names, impact_values, color=colors)
+        plt.title("🧠 Feature Influence on LSTM Prediction")
+        plt.xlabel("Impact on Price")
+        for bar, value in zip(bars, impact_values):
+            plt.text(bar.get_width(), bar.get_y() + bar.get_height()/2, f"{value:.4f}", va='center', ha='left', fontsize=9)
         plt.tight_layout()
         buf_shap = io.BytesIO()
         plt.savefig(buf_shap, format="png")
@@ -467,13 +475,14 @@ def predict_price(symbol):
             tech_summary,
             corp_info,
             sector_chart,
-            shap_chart  # NEW output!
+            shap_chart
         )
 
     except Exception as e:
         blank = PIL.Image.new("RGB", (400, 200))
         return ("Error", "-", str(e), blank, blank, blank, blank, "No news", "No news", {}, "", "", blank, blank)
 
+
 def simulate_portfolio(portfolio_str):
     try:
         entries = [e.strip() for e in portfolio_str.split(',')]