Skip to content

LSTM Model for Crypto Trading Volume Prediction

Introduction to LSTM

Long Short-Term Memory (LSTM) networks are a type of recurrent neural network (RNN) architecture designed to model temporal sequences and their long-range dependencies more accurately than conventional RNNs. In our project, we use LSTM to predict cryptocurrency trading volumes based on historical data.

Model Architecture

Our LSTM model consists of the following layers:

  1. LSTM Layer: This is the core of our model. It processes the input sequence and captures temporal dependencies.
  2. Units: 50
  3. Activation: ReLU (Rectified Linear Unit)

  4. Input Shape: (look_back, 1), where look_back is the number of past time steps to consider

  5. Dense Layer: This is the output layer that produces the final prediction.

  6. Units: 1 (since we're predicting a single value - the trading volume)

Model Parameters

  • look_back: This parameter determines how many past time steps the model considers when making a prediction. It's set in the config.yml file and can be adjusted based on the specific requirements of the analysis.

Training Process

The model is trained using the following settings:

  • Optimizer: Adam
  • Loss Function: Mean Squared Error (MSE)
  • Epochs: 100 (default, can be adjusted)
  • Batch Size: 32 (default, can be adjusted)

Input Data

The model expects input data in the shape (n_samples, look_back, 1), where: - n_samples is the number of training examples - look_back is the number of past time steps to consider - 1 represents the single feature we're using (volume)

Output

The model outputs a single value representing the predicted trading volume for the next time step.

Potential Improvements

  1. Feature Engineering: Incorporate additional features such as price data, moving averages, or technical indicators.
  2. Hyperparameter Tuning: Use techniques like grid search or random search to optimize the model's hyperparameters.
  3. Ensemble Methods: Combine the LSTM model with other types of models to potentially improve prediction accuracy.
  4. Attention Mechanism: Implement an attention layer to help the model focus on the most relevant parts of the input sequence.

For implementation details, please refer to the model.py file in the project root directory.