afdemir06/finance-predict
Folders and files
| Name | Name | Last commit date | ||
|---|---|---|---|---|
Repository files navigation
# AI-Powered Stock Predictor An end-to-end machine learning application that predicts tomorrow's share price direction based on a user-defined historical period. Unlike static models, this app trains a new Random Forest model for any ticker symbol (e.g. `BTC-USD`, `TSLA`), making it fully dynamic and adaptable. --- ## Features ### Automated Update Logic Models are automatically retrained if the saved weights are older than 7 days, ensuring the AI stays synchronized with the most recent market data. ### Advanced Feature Engineering Automatically computes technical indicators including: - **Relative Strength Index (RSI)** - **Bollinger Bands** (Upper / Lower) - **Volatility** (Rolling standard deviation) - **Price Momentum & Lagged Returns** ### Confidence Thresholding Implements a probability filter — a "buy" signal is only triggered when the model's confidence exceeds **60% probability** — to reduce noise and false positives. ### Performance Optimized Integrated with **Streamlit Cache** (TTL: 1 hour) to minimize redundant API calls to Yahoo Finance. --- ## Usage 1. Enter a stock or crypto symbol in the sidebar (e.g. `AAPL`, `BTC-USD`) 2. Select the historical period for training data 3. The model will train automatically (or click **Force Train** to retrain manually) 4. View the prediction and price chart