Home Arrow Icon Knowledge base Arrow Icon Global Arrow Icon How can I use the Darts library to compare different RNN models


How can I use the Darts library to compare different RNN models


To compare different RNN models using the Darts library in Python, you can follow these steps:

1. Install Darts

- Install Darts using pip:
bash
     pip install darts
     

2. Import Libraries

- Import necessary libraries:
python
     import darts
     import pandas as pd
     import numpy as np
     

3. Prepare Data

- Load your time series data into a pandas DataFrame:
python
     df = pd.read_csv('your_data.csv')
     

4. Split Data

- Split the data into training and testing sets:
python
     from sklearn.model_selection import train_test_split
     train, test = train_test_split(df, test_size=0.2, random_state=42)
     

5. Create Darts Time Series

- Create Darts TimeSeries objects for the training and testing data:
python
     from darts import TimeSeries
     train_series = TimeSeries(train)
     test_series = TimeSeries(test)
     

6. Create RNN Models

- Create instances of the RNN models you want to compare:
python
     from darts.models import RNNModel, BlockRNNModel

     # LSTM Model
     lstm_model = RNNModel(model="LSTM", hidden_dim=20, dropout=0.2, batch_size=16, n_epochs=300)

     # GRU Model
     gru_model = RNNModel(model="GRU", hidden_dim=20, dropout=0.2, batch_size=16, n_epochs=300)

     # Vanilla RNN Model
     vanilla_model = RNNModel(model="Vanilla", hidden_dim=20, dropout=0.2, batch_size=16, n_epochs=300)
     

7. Train Models

- Train each model on the training data:
python
     lstm_model.fit(train_series)
     gru_model.fit(train_series)
     vanilla_model.fit(train_series)
     

8. Evaluate Models

- Evaluate each model on the testing data:
python
     from darts.metrics import mape

     lstm_mape = mape(lstm_model.predict(test_series), test_series)
     gru_mape = mape(gru_model.predict(test_series), test_series)
     vanilla_mape = mape(vanilla_model.predict(test_series), test_series)

     print("LSTM MAPE:", lstm_mape)
     print("GRU MAPE:", gru_mape)
     print("Vanilla RNN MAPE:", vanilla_mape)
     

9. Plot Results

- Plot the actual values and the predictions from each model:
python
     import matplotlib.pyplot as plt

     plt.figure(figsize=(12, 6))

     plt.plot(test_series.values, label='Actual')
     plt.plot(lstm_model.predict(test_series).values, label='LSTM')
     plt.plot(gru_model.predict(test_series).values, label='GRU')
     plt.plot(vanilla_model.predict(test_series).values, label='Vanilla RNN')

     plt.legend()
     plt.show()
     

Conclusion

- By following these steps, you can compare different RNN models using the Darts library in Python. This allows you to evaluate their performance and choose the best model for your time series forecasting task[1][2][4].
Citations:
[1] https://unit8co.github.io/darts/examples/04-RNN-examples.html
[2] https://www.kaggle.com/code/ferdinandberr/darts-forecasting-deep-learning-global-models
[3] https://towardsdatascience.com/temporal-loops-intro-to-recurrent-neural-networks-for-time-series-forecasting-in-python-b0398963dc1f?gi=ca1e8d29f8ab
[4] https://unit8.com/resources/training-forecasting-models/
[5] https://unit8co.github.io/darts/