In this video, we look at some approaches that could not give us high AUC scores. I also share my opinion about why they don't work really well. After editing this video, I feel like at least one of the lessons might sound like an obvious lesson. But, I guess, we can just forget about it sometimes when we are searching for the best models.
Thank you so much for watching!
00:00 Intro
00:44 Description of the data and data processing
04:38 Approach 1: Shapelets and gradient boosting
12:42 Approach 2: Shapelets and logistic regression
13:50 Approach 3: Shapelets and neural network
15:48 Approach 4: LSTM
22:11 Approach 5: LSTM with more features
27:58 Takeaways and a question
Source code: https://github.com/stephanielees/time-series-classification-on-sensor-data/blob/main/models_with_low_AUC.ipynb
#timeseries #classification #neuralnetworks #datascience #datamining #deeplearning #python #kaggle