ptw_ti_classification

Classification task from PTW Lab

Problem statement

Data sets

Approach

Setup

Update log

• 04.08.2022 Tried features from tsfresh again, the lgbm can not identify the negative sample at in on second data set.
  • Failed in selected feature from tsfresh. The selected feature(~6500) from first data set and second data set(~1200) are not exactly same.
• 03.08.2022 used lgbm model on fft data, achieve 0.88 on second data set
  • The extracted feature from tsfresh didn't show good f1-score. We might need feature selection and fine-tuning the lgbm model.
• 03.08.2022 rewrite the pipeline in train.py

  • This will make the code more readable.

• 01.08 create the repository