In this section, we will examine in more detail the datasets that we will be using:
- HAR dataset: The dataset is a text file that consists of the different subjects accelerations for each of the six activities. We can do a data distribution check for the dataset as it is not easy to perceive the data distribution by looking at the text file only. The following graph summarizes the breakdown for the training set:
As we can see from the preceding graph, the training dataset consists of more walking and jogging data than the other four activities. This is good for the DL model, since walking and jogging are moving activities, where the range of acceleration data could be wide. To visualize this, we have explored activity-wise acceleration measurements/data for 200 time steps for each activity. The following screenshot represents 200 time step acceleration measurements...
