📈 YFinance Analyzer

Predict financial asset prices using deep learning — powered by LSTM neural networks.

Quick Start · Features · How It Works · Example · License

---

YFinance Analyzer is a command-line tool that fetches real-time market data, engineers technical indicators, trains an LSTM model on historical prices, and produces a clean forecast chart — all in a single script. Just pick a ticker, give it a date range, and let the model do the rest.

🚀 Quick Start

Prerequisites

• Python 3.10+

Installation

# Clone the repository
git clone https://github.com/Azmvv/YFinance-Analyzer.git
cd YFinance-Analyzer

# Install dependencies
pip install -r requirements.txt

Usage

python3 predictor.py

You'll be prompted for three inputs:

Prompt	Example	Description
Ticker symbol	BTC-USD, AAPL, TSLA	Any valid Yahoo Finance ticker
Start date	2020-01-01	Beginning of the historical window
End date	2025-12-31	End of the historical window

The script handles everything from there — data download, feature engineering, model training, evaluation, and chart generation.

✨ Features

• Any Yahoo Finance asset — stocks, crypto, ETFs, indices, forex pairs, and more.
• Technical indicators — automatically computes a 14-day RSI and 7-day Moving Average to give the model richer context.
• Sliding-window sequences — uses the past 60 days of multi-feature data to predict the next day's closing price.
• Two-layer LSTM with Dropout — deep enough to capture temporal patterns, regularized to avoid overfitting.
• Early stopping — training halts automatically when validation loss stops improving, saving time and preventing overfit.
• 7-day future forecast — after evaluation, the model rolls forward to predict the next 7 business days beyond your dataset.
• Publication-quality chart — a single PNG with actual prices, test-set predictions, and the forecast line, complete with legend, grid, and formatted dates.

⚙️ How It Works

User Input           Data Pipeline              Model                  Output
───────────     ─────────────────────     ─────────────────     ─────────────────
 Ticker    ──▶  yfinance download    ──▶  LSTM (128 → 64)  ──▶  Test predictions
 Start date     + RSI, 7-day MA            + Dropout 0.3         Future forecast
 End date       + MinMaxScaler             + EarlyStopping        Saved PNG chart
                + 60-day windows           + Adam / MSE           MAE & MAPE

Step-by-step

1. Fetch — Historical OHLCV data is downloaded from Yahoo Finance via the yfinance library.
2. Engineer — Two technical indicators are computed and appended as new columns:
   • 7-day Moving Average — smooths short-term price noise.
   • 14-day RSI — measures momentum (overbought/oversold conditions).
3. Scale — All features (Close, Volume, MA_7, RSI_14) are normalized to [0, 1] using MinMaxScaler.
4. Window — Data is reshaped into overlapping 60-day sequences, each paired with the next day's closing price as the label.
5. Split — 80% of sequences go to training, 20% to testing (chronological, no shuffling).
6. Train — A two-layer LSTM network trains for up to 50 epochs with early stopping (patience = 8).
7. Evaluate — The model predicts on the held-out test set. MAE and MAPE are printed to the console.
8. Forecast — The last 60-day window is fed back into the model iteratively to generate 7 future price points.
9. Plot — Everything is visualized on a single matplotlib chart and saved as a high-resolution PNG.

Model Architecture

Layer               Output Shape         Parameters
─────────────────────────────────────────────────────
LSTM (128 units)    (batch, 60, 128)     68,096
Dropout (0.3)       (batch, 60, 128)     0
LSTM (64 units)     (batch, 64)          49,408
Dropout (0.3)       (batch, 64)          0
Dense (32, ReLU)    (batch, 32)          2,080
Dense (1, linear)   (batch, 1)           33
─────────────────────────────────────────────────────
Total trainable parameters: ~119,617

📊 Example

$ python3 predictor.py
Enter ticker symbol (e.g. BTC-USD, AAPL): AAPL
Enter start date (YYYY-MM-DD): 2020-01-01
Enter end date   (YYYY-MM-DD): 2025-12-31

📥  Downloading AAPL data from 2020-01-01 to 2025-12-31 …
✅  Received 1,487 rows after feature engineering.

🔢  Training samples : 1,142
🔢  Testing samples  : 286

🚀  Training LSTM model …
Epoch 1/50 ━━━━━━━━━━━━━━━━━━ 33/33 - loss: 0.0042 - val_loss: 0.0018
...
Epoch 19/50 ━━━━━━━━━━━━━━━━━━ 33/33 - loss: 0.0005 - val_loss: 0.0004

📈  Test MAE  : 4.32
📈  Test MAPE : 1.87%

🔮  7-Day Forecast from 2026-01-02:
    2026-01-02  →  $248.71
    2026-01-03  →  $249.15
    ...

A chart like this is saved automatically:

AAPL_prediction.png — Actual vs. Predicted prices with a 7-day forecast extension.

🗂 Project Structure

YFinance-Analyzer/
├── predictor.py        # Main script — data, model, training, plotting
├── requirements.txt    # Python dependencies
├── LICENSE             # AGPL-3.0
└── README.md           # You are here

📦 Dependencies

Package	Purpose
numpy	Numerical operations
pandas	Data manipulation & time series
yfinance	Yahoo Finance market data API
scikit-learn	MinMaxScaler for feature normalization
tensorflow	LSTM model (Keras API)
matplotlib	Chart generation

⚠️ Disclaimer

This tool is built for educational and research purposes only. Financial markets are inherently unpredictable — no model can guarantee future returns. Do not use these predictions as the sole basis for real trading or investment decisions. Always do your own research and consult a qualified financial advisor.

📄 License

This project is licensed under the GNU Affero General Public License v3.0 — see the LICENSE file for details.