This repository has various use-cases of activity recognition, applying different ML techniques for classification-tasks.
has a set of different scenarios that apply different transfer-learning techniques in activity-recognition datasets
is a collection of projects that apply this technique in HAR
is a collection of projects that apply this technique in different scenarios besides HAR
handles the pre-processing of the datasets
The construction of the model must take into account the dependency between the collected signals, since they consist in a time-series.
Therefore avoid typical cross-validation and apply techniques such as forward-chaining, leave-one-subject-out, cross-validation for time-series.
see more here:
https://stats.stackexchange.com/questions/14099/using-k-fold-cross-validation-for-time-series-model-selection
https://robjhyndman.com/hyndsight/tscv/