📈 Stock Price Prediction Using LSTM

This project demonstrates how to use Long Short-Term Memory (LSTM), a type of Recurrent Neural Network (RNN), to predict stock prices based on historical data. It includes steps for data preprocessing, sequence generation, LSTM model training, and performance visualization.

---

🛠️ Features

• Loads historical stock data
• Normalizes data using MinMaxScaler
• Prepares input sequences with configurable time_step
• Builds and trains an LSTM-based neural network
• Predicts future stock prices
• Visualizes actual vs. predicted prices

---

📂 Project Structure

Stock_Price_prediction_LSTM.ipynb
README.md

---

📦 Requirements

Install the required packages using:

pip install numpy pandas matplotlib scikit-learn tensorflow

---

📊 How It Works

1. Data Preprocessing

   • Loads stock price data (e.g., 'Close' prices).
   • Normalizes using MinMaxScaler to scale values between 0 and 1.

2. Sequence Generation

   • Uses a prepare_data() function that creates input-output pairs based on a given time_step (default: 60).

3. LSTM Model

   • A sequential Keras model with stacked LSTM layers and Dense output.
   • Trained on historical sequences to predict future stock prices.

4. Evaluation & Visualization

   • Predicts on test data.
   • Visualizes actual vs. predicted prices using Matplotlib.

---

⚙️ Key Function: prepare_data()

def prepare_data(data, time_step=60):
    scaler = MinMaxScaler(feature_range=(0, 1))
    data_scaled = scaler.fit_transform(np.array(data).reshape(-1, 1))
    X, y = [], []
    for i in range(time_step, len(data_scaled)):
        X.append(data_scaled[i-time_step:i, 0])
        y.append(data_scaled[i, 0])
    return np.array(X), np.array(y), scaler

This function generates sequences and targets from time series data, ready for LSTM input.

---

📉 Sample Output

• Graph comparing actual vs predicted prices
• Evaluation metrics (like RMSE, optional)

---