Log 1: 13th September

[io23632]

General Notes:

Feature Engineering tutorial

• From my meeting with Jon last time I used the feature engineering tutorial to https://towardsdatascience.com/feature-engineering-on-time-series-data-transforming-signal-data-of-a-smartphone-accelerometer-for-72cbe34b8a60 and followed along to feature extraction of 36 users accelerometer data.
• Here I used LazyPredict as well as the Logistic Regression to predict the user's activity based on the various features extracted. And found that a model called ExtraTreeClassiffer had low computation and a higher accuracy score then the Logistic Regression Model.
• I then had a meeting with Jon to disscuss these results.

Today's Plan:

Personalised Model

1. Research K fold validation
2. Make personalised model data from the WIDMS data set
3. Using k fold validation train various classification models on personalised data of a single user
4. Repeat with all the 36 users data
5. Compare model accuracy of personalised data with impersonalised data