initial commit

.github/workflows/sync-to-hf-space.yml

@@ -0,0 +1,40 @@
+# .github/workflows/sync-to-hf-space.yml
+name: Sync to Hugging Face Space
+
+on:
+  push:
+    branches:
+      - main # Or whatever your default branch is
+
+jobs:
+  sync:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v3
+        with:
+          fetch-depth: 0 # Fetch all history for accurate diffs
+
+      - name: Push to Hugging Face Space
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN }}
+          # Your Hugging Face username
+          HF_USERNAME: GuitarGeorge
+          # The ID of your Space (username/space-name)
+          SPACE_ID: GuitarGeorge/Neural-Prophet-Predictor
+        run: |
+          echo "Attempting to push to Hugging Face Space: $SPACE_ID"
+          # Add Hugging Face as a remote
+          git remote add hf_space "https://dummy:${HF_TOKEN}@huggingface.co/spaces/${SPACE_ID}"
+          
+          # Check if the remote was added
+          git remote -v
+          
+          # Configure Git user
+          git config --global user.email "action@github.com"
+          git config --global user.name "GitHub Action"
+          
+          # Force push to ensure the Space mirrors GitHub
+          git push --force hf_space HEAD:refs/heads/main
+          
+          echo "Push to Hugging Face Space completed."

README.md

@@ -0,0 +1,58 @@
+---
+title: Neural Prophet Stock Predictor
+emoji: 🔮
+colorFrom: indigo
+colorTo: blue
+sdk: gradio
+sdk_version: 5.29.1
+app_file: app.py
+python_version: 3.10
+pinned: false
+---
+
+# NeuralProphet Stock Predictor 📈
+
+## Project Goal
+This project uses **NeuralProphet**, a PyTorch-based hybrid forecasting framework (inspired by Facebook Prophet and Neural Networks), to predict the **90-day forward price** for any valid stock ticker found on Yahoo Finance. 
+
+Unlike simple linear regressions, this model accounts for time-series specific components such as:
+- **Trend Changes:** Detecting shifts in the stock's growth trajectory.
+- **Seasonality:** analyzing Yearly and Weekly patterns (e.g., does the stock typically rise in January?).
+
+## How to Use
+
+### Using the UI
+1. Enter a valid Ticker Symbol in the textbox (e.g., `AAPL` for Apple, `AZN.L` for AstraZeneca UK).
+2. Click **"Analyze Stock"**.
+3. Wait approximately 10-30 seconds for the model to train on-the-fly.
+4. View the **ROI Verdict**, the **Forecast Chart**, and the **Seasonality breakdown**.
+
+### Interpretation
+- **Blue Line (Forecast):** The predicted price path.
+- **Yearly Seasonality:** Shows which months are historically bullish or bearish for this specific asset.
+- **Verdict:**
+    - 🟢 **STRONG BUY:** ROI > 10%
+    - 🟢 **BUY:** ROI > 2%
+    - 🟡 **HOLD:** ROI > -5%
+    - 🔴 **SELL:** ROI <= -5%
+
+## Model Methodology
+
+This application trains a *fresh* model every time you request a ticker. It does not use pre-trained weights because stock data changes daily.
+
+1.  **Data Fetching:** Downloads 3 years of daily historical data via `yfinance`.
+2.  **Preprocessing:** Cleans data and handles timezone formatting.
+3.  **Training:** Fits a NeuralProphet model with:
+    *   Yearly Seasonality: Enabled
+    *   Weekly Seasonality: Enabled
+    *   Daily Seasonality: Disabled (Noise reduction)
+    *   Learning Rate: 0.01
+4.  **Forecasting:** Projects 90 days into the future.
+
+## Tech Stack
+- **NeuralProphet:** For time-series forecasting.
+- **YFinance:** For live market data.
+- **Gradio:** For the web interface.
+- **Plotly:** For interactive charting.
+
+*Disclaimer: Stock market predictions are inherently uncertain. Do not trade based solely on these results.*

app.py

@@ -0,0 +1,154 @@
+import gradio as gr
+import torch
+import logging
+import warnings
+import os
+import yfinance as yf
+import pandas as pd
+from neuralprophet import NeuralProphet
+import plotly.graph_objs as go
+
+# --- STEP 1: CONFIGURATION & PATCHES ---
+
+# Suppress messy logs
+logging.getLogger("neuralprophet").setLevel(logging.ERROR)
+logging.getLogger("pytorch_lightning").setLevel(logging.ERROR)
+warnings.filterwarnings("ignore")
+os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
+
+# Fix for PyTorch 2.6+ security check
+original_load = torch.load
+def patched_load(*args, **kwargs):
+    if 'weights_only' not in kwargs:
+        kwargs['weights_only'] = False
+    return original_load(*args, **kwargs)
+torch.load = patched_load
+
+# --- STEP 2: PREDICTION LOGIC ---
+
+def predict_stock(ticker):
+    """
+    Takes a ticker symbol, trains a NeuralProphet model, 
+    and returns a textual report and two Plotly figures.
+    """
+    ticker = ticker.strip().upper()
+    
+    if not ticker:
+        return "⚠️ Please enter a ticker symbol.", None, None
+
+    status_msg = f"⏳ Analyzing {ticker}... Please wait (approx 10-30 seconds)."
+    print(f"Processing {ticker}")
+
+    try:
+        # 1. Get Data
+        data = yf.download(ticker, period="3y", interval="1d", progress=False)
+
+        # Handle cases where yfinance returns empty dataframe or multi-index columns
+        if data.empty:
+            return f"❌ Could not find data for ticker '{ticker}'. Please check the symbol.", None, None
+        
+        # Flatten MultiIndex if present (yfinance update quirk)
+        if isinstance(data.columns, pd.MultiIndex):
+            try:
+                # Attempt to extract just the Close column for the specific ticker
+                df = data.xs(ticker, axis=1, level=1)
+                if 'Close' in df.columns:
+                    df = df[['Close']].reset_index()
+                else:
+                    # Fallback for single level extraction
+                    df = data['Close'].reset_index()
+            except:
+                # Brute force flatten
+                df = data.copy()
+                df.columns = ['_'.join(col).strip() for col in df.columns.values]
+                # Look for a column containing "Close"
+                close_col = [c for c in df.columns if "Close" in c][0]
+                df = df[[close_col]].reset_index()
+        else:
+            df = data[['Close']].reset_index()
+
+        # Rename for NeuralProphet
+        df.columns = ['ds', 'y']
+        
+        # Ensure dates are timezone-naive
+        df['ds'] = df['ds'].dt.tz_localize(None)
+
+        if len(df) < 100:
+            return f"❌ Not enough historical data found for {ticker} (Need > 100 days).", None, None
+
+        # 2. Train Model
+        m = NeuralProphet(
+            yearly_seasonality=True,
+            weekly_seasonality=True,
+            daily_seasonality=False,
+            learning_rate=0.01,
+            # Disable progress bars for clean logs
+            trainer_config={"enable_progress_bar": False} 
+        )
+
+        m.fit(df, freq="D")
+
+        # 3. Predict 90 Days out
+        future = m.make_future_dataframe(df, periods=90)
+        forecast = m.predict(future)
+
+        # 4. Extract Metrics
+        current_price = df['y'].iloc[-1]
+        predicted_price = forecast['yhat1'].iloc[-1]
+
+        # Calculate ROI
+        roi = ((predicted_price - current_price) / current_price) * 100
+
+        # Generate Verdict
+        if roi > 10: verdict = "STRONG BUY 🟢"
+        elif roi > 2: verdict = "BUY 🟢"
+        elif roi > -5: verdict = "HOLD 🟡"
+        else: verdict = "SELL 🔴"
+
+        # 5. formatting Output Text
+        report = f"""
+### 📊 Analysis Report: {ticker}
+**Current Price:** {current_price:.2f}
+**90-Day Target:** {predicted_price:.2f}
+**Projected ROI:** {roi:.2f}%
+**Verdict:** {verdict}
+
+*Disclaimer: This is an AI-generated forecast based on historical trends. Not financial advice.*
+        """
+
+        # 6. Generate Plots
+        # NeuralProphet plot() returns a plotly figure object
+        fig_forecast = m.plot(forecast)
+        fig_components = m.plot_components(forecast)
+
+        return report, fig_forecast, fig_components
+
+    except Exception as e:
+        import traceback
+        traceback.print_exc()
+        return f"❌ An error occurred while processing {ticker}: {str(e)}", None, None
+
+# --- STEP 3: GRADIO INTERFACE ---
+
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    gr.Markdown("# 📈 NeuralProphet Stock Predictor")
+    gr.Markdown("Enter a stock ticker (e.g., `AAPL`, `TSLA`, `AZN.L`) to generate a 90-day forecast.")
+    
+    with gr.Row():
+        ticker_input = gr.Textbox(label="Ticker Symbol", placeholder="e.g. AZN.L", value="AZN.L")
+        submit_btn = gr.Button("Analyze Stock", variant="primary")
+    
+    result_text = gr.Markdown(label="Verdict")
+    
+    with gr.Row():
+        plot1 = gr.Plot(label="Price Forecast")
+        plot2 = gr.Plot(label="Seasonality Components")
+
+    submit_btn.click(
+        fn=predict_stock, 
+        inputs=ticker_input, 
+        outputs=[result_text, plot1, plot2]
+    )
+
+if __name__ == "__main__":
+    demo.launch()

readme.md

@@ -1 +0,0 @@
-

requirements.txt

@@ -0,0 +1,6 @@
+neuralprophet
+yfinance
+pandas
+gradio
+plotly
+torch