CAERise Dataset

How we built it

We did several handcrafts and deep learning

Challenges we ran into

Time series analysis, under 500 samples, risk of mislabel data, unbalanced data and difference data size.

Accomplishments that we're proud of

There are five challenges in total and we came up with solutions with each of them: Time series analysis LSTM (long short term memory) and CNN (convolution neural network) Under 500 samples Windowing Risk of mislabel data We didn’t have time to do this, but in order to tackle this problem, we can use weak supervised learning Unbalanced data Scoring matrix, F1 score and area under curve Difference data size Measurement of cross correlation and windowing

What's next for CAERise Dataset

Try some preprocessing method
Try other feature extraction and reduction

Note

Team Name: Data Pirates

Team Members: Abderrahim Khalifa (Polytechnique Montreal, Master student, khalifa.abde95@gmail.com) Mohammad Javad Darvishi (UDEM, PHD student, 653mjd@gmail.com) Suzy Liu (McGill, Undergraduate student, suzylxx@gmail.com)

Project Name: CAERise Dataset

GitHub Link: https://github.com/MohammadJavadD/ImpAI2019

Documentation Link: https://docs.google.com/document/d/1Zb2WzPsGEl38caHymcJZHS6xF9ZEauKKg8jb4ZPQtqY/edit?usp=sharing

Presentation Link: https://docs.google.com/presentation/d/1QXF8dv90jYIqdmEyEdziblRFX8mlVLaj_7R6JBYua7c/edit?usp=sharing