The original Data link: Dataset

Dataset created from raw data of co-ordinates from both accelerometer and gyroscope from both mobile and smart watch.

In Preprocessing of data rather than using multiple values I too mean, variance, max and min of the X,Y and Z co-ordinates

I have implemented the problem in three different ways

  1. using only watch data 
  2. using only phone data 
  3. using both watch and mobile data

The max accuracy using machine learning model is 66% and it was achieved using only watch data for the models