https://github.com/hayatiyrtgl/cryptocurrency_time_series_rnn

Python script for training a Simple RNN model on cryptocurrency price data to predict future prices, including data exploration and evaluation
https://github.com/hayatiyrtgl/cryptocurrency_time_series_rnn

data-analysis data-science data-visualization keras pandas pandas-python prediction predictive-modeling python python-script rnn rnn-tensorflow tensorflow time-series time-series-analysis

Last synced: 3 days ago
JSON representation

Python script for training a Simple RNN model on cryptocurrency price data to predict future prices, including data exploration and evaluation

• Host: GitHub
• URL: https://github.com/hayatiyrtgl/cryptocurrency_time_series_rnn
• Owner: HayatiYrtgl
• License: mit
• Created: 2024-04-25T07:51:03.000Z (7 months ago)
• Default Branch: main
• Last Pushed: 2024-04-25T07:52:13.000Z (7 months ago)
• Last Synced: 2024-04-25T08:46:53.246Z (7 months ago)
• Topics: data-analysis, data-science, data-visualization, keras, pandas, pandas-python, prediction, predictive-modeling, python, python-script, rnn, rnn-tensorflow, tensorflow, time-series, time-series-analysis
• Language: Jupyter Notebook
• Homepage:
• Size: 0 Bytes
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
This code is a Python script for building and training a Simple Recurrent Neural Network (RNN) using Keras to predict cryptocurrency prices. Here's a breakdown of what each section does:

1. **Importing Libraries**: This section imports necessary libraries such as pandas, numpy, scikit-learn, keras, and matplotlib.

2. **Reading Data**: Reads cryptocurrency price data from a CSV file.

3. **Data Exploration**:

   - Checks for missing values.

   - Displays data types and descriptive statistics.

   - Visualizes data using scatter plots and a heatmap of correlations between features.

4. **Data Preprocessing**:

   - Converts the timestamp column to datetime format.

   - Sets the timestamp column as the index.

5. **Train-Test Split**:

   - Manually splits the data into training and testing sets.

   - Defines a function `train_test_split` to create input-output pairs for training the model.

6. **Creating the RNN Model**:

   - Uses Keras Sequential API to define an RNN model.

   - The model consists of an Input layer, a SimpleRNN layer with 60 units and 'tanh' activation function, and a Dense output layer.

   - Compiles the model using mean squared error loss and Adam optimizer, with R2Score as a metric.

7. **Training the Model**:

   - Fits the model to the training data with early stopping and model checkpointing callbacks.

   - Validates the model on a validation split of the training data.

8. **Evaluation**:

   - Predicts cryptocurrency prices using the trained model.

   - Plots the predicted prices against the true prices for evaluation.

Overall, this script demonstrates a basic workflow for building and training a recurrent neural network for cryptocurrency price prediction using historical data.