Activity Prediction

Classification of time-series recorded by smartphone sensors (e.g. human activity prediction using the accelerometer data).

There are two parties: a server and a client, which communicate via the asynchronous TCP connection.

The client collects data using the smartphone sensors and sends it to the server for:

• time-series classification,
• training a classifier.

The back end (the server) receives the data and performs classification remotely thereby allowing for using complex classification algorithms.

The server works as follows. It saves the received time-series, performs computations, and responds to the client with a predicted class label, if necessary.

All the data received by the server is logged on the server in directories specified by clients (the ‘account’ field in the client app).

The server was implemented in Python, the client in Java.

The source code is open and is available from the project repository: https://github.com/karasikov/TimeSeriesClassification.