trademind-lstm

TradeMind LSTM Stock Predictor

LSTM-based stock price forecasting model trained on 2 years of daily OHLCV data.

Model Details

• Architecture: Stacked LSTM (128 → 64 → 32 units) with BatchNorm + Dropout
• Framework: TensorFlow / Keras
• Input: 60-day sliding window of 7 features
• Output: 5-day price forecast
• Features: Close price, Returns, MA5, MA20, Volatility, RSI(14), MACD

Trained Symbols

AAPL, TSLA, NVDA, MSFT, AMZN, GOOGL, META, AMD, JPM, NFLX

Usage

import pickle
import pandas as pd
from tensorflow.keras.models import load_model
import numpy as np

# Load model and scaler
model = load_model("AAPL_lstm.keras")
with open("AAPL_scaler.pkl", "rb") as f:
    scaler = pickle.load(f)

# Prepare your DataFrame with columns: close, returns, ma5, ma20, volatility, rsi, macd
# df = your_dataframe  (at least 60 rows)

features = ["close", "returns", "ma5", "ma20", "volatility", "rsi", "macd"]
data = df[features].values
scaled = scaler.transform(data)
seq = scaled[-60:].reshape(1, 60, len(features))

pred_scaled = model.predict(seq)[0]
dummy = np.zeros((5, len(features)))
dummy[:, 0] = pred_scaled
predicted_prices = scaler.inverse_transform(dummy)[:, 0]
print("5-day forecast:", predicted_prices)

Disclaimer

This model is for educational and research purposes only. Not financial advice. Past performance does not guarantee future results.