Skip to content

An activity recognizer using a neural network implemented from scratch

Notifications You must be signed in to change notification settings

vesran/Human-Activity-Recognition

Repository files navigation

Human-Activity-Recognition

Classify different physical activities from motion sensors data (accelerometers and gyroscope) colleted in smartphones and smartwatches by using a multilayer perceptron (implemented from scratch). Activities : ‘Biking’, ‘Sitting’, ‘Standing’, ‘Walking’, ‘Stair Up’ and ‘Stair down’.

Average_error% = 18% (k-fold cross-validation)

Dataset

Link: https://archive.ics.uci.edu/ml/datasets/Heterogeneity+Activity+Recognition

561 columns ; 7352 lines

There two main files : one contains the data without classes and one contains classes only.

Ackknowledgement

This personal project was intended as an exercise to understand the concept of Neural Network in Machine Learning. That is the reason why I chose to implement everything from scratch.