This package is implemented by Sharareh Sayyad. It explores the classification problem for a time-dependent dataset. In particular, the code is implemented to address this problem on WISDM dataset.

Various deep learning models are implemented in this package, including:

• MLP: The codes are stored in MLP/
• CNN: The codes are stored in CNN_LSTM/
• LSTM: The codes are stored in CNN_LSTM/
• CNN-LSTM: The codes are stored in CNN_LSTM/
• Transformer: The codes are stored in Transformer/

A notebook to analyze the statistical behavior of the dataset is available here.

A collection of required packages to install and execute our code is stored in requirements.txt.

To install the package, one can either call sh install.sh or try pip install . inside the package's main directory.

After installing the package, all provided notebooks and Python codes in data_analysis, MLP, CNN_LSTM and Transformer directories can be used.

See the report file for further details.

I have collected a selected number of my best performed parameter sets and their associated metrics on the test sets in postprocessing_avg/. The results are obtained from averaging the outputs of each particular model for 20 different trainings.