Stock-Price-Predictor

Imported from OmSwainSupreme/Stock-Price-Predictor (AGENTS.md). Install upstream with npx skills add OmSwainSupreme/Stock-Price-Predictor. Copyright stays with the author.

AGENTS.md - Agent Instructions for Stock Price Predictor

Project Overview

This is a complete Stock Price Predictor application built with Streamlit, designed for AI/ML internship demonstration. The application retrieves historical stock data, engineers features, trains a Linear Regression model with chronological splitting, evaluates on holdout data, and predicts the next trading session's close.

Key Architectural Decisions

1. No Temporal Leakage

• All features computed with explicit shifts (.shift(1) for lags, rolling windows shifted)
• Target is t+1 close, features from t, t-1, ...
• Chronological train/test split (80/20) - NO random shuffling
• Final deployment model trained on ALL data AFTER evaluation

2. Data Flow

yfinance -> normalize_data() -> create_features() -> chronological_split() 
-> fit(train) -> predict(test) -> evaluate -> fit(all) -> predict_next()

3. Price Series

• Uses Adjusted Close (accounts for splits/dividends) when available
• Falls back to Close if Adj Close unavailable
• Documents which series is used throughout

4. Error Handling

• Validates ticker format, data sufficiency, non-finite values
• Graceful degradation with user-friendly messages
• Never fabricates data or silently uses stale cache

Module Responsibilities

Testing Strategy

• Unit tests only - no internet, no live data
• Synthetic deterministic data (seeded random walk)
• Test categories:
  • Feature column generation & alignment
  • No future leakage in rolling features
  • Chronological split preserves order
  • Train/test separation enforced
  • Metrics calculations (including edge cases)
  • Next-session feature reconstruction matches training
  • Error handling for invalid/empty inputs

Running Commands (Verified)

# Windows - Setup
cd C:\codecproject1
python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt

# Run tests
pytest tests/ -v

# Run application
streamlit run app.py

Common Issues & Fixes

Extension Points

1. Add models: Extend model.py with new estimator classes (keep same interface)
2. Add features: Modify features.py FEATURE_COLUMNS and create_features()
3. Add metrics: Extend metrics.py calculate_metrics()
4. Change data source: Replace data_loader.py fetch logic
5. Add charts: Add Plotly figures in app.py tabs

Do Not

• ❌ Add random train/test split
• ❌ Use future data in features (no shift(-1) except for target)
• ❌ Train on test data
• ❌ Fabricate confidence intervals
• ❌ Claim financial accuracy
• ❌ Add deep learning without clear need
• ❌ Commit virtual environments or cache

Git Workflow

git init
git add .
git commit -m "feat: implement stock price prediction dashboard"
# Push to remote if needed