These files contain scripts to extract features from motion data and evaluate models for a user a user authorization system. The data are from 15 human subjects performing a variety of tasks taken from the University of California Irvine Machine Learning repository. Time and frequency domain features are extracted and used in various classifiers to predict an authorized user or unauthorized user (i.e. any other user). An average F1 score over all users is computed to evaluate model performance.
The index labels in the data appear to have alignment issues. Run
python fix_ind.py
to create new files with adjusted indidies. The load_file function will use the indicies in these files if the files exist.
To run the analysis run:
python analysis.py
Most functions, however, will have increased functionability if run from an interpreter.
Python 2.7 scikit-learn 0.17.1
Data files can be found at: http://archive.ics.uci.edu/ml/datasets/Activity+Recognition+from+Single+Chest-Mounted+Accelerometer
This project contains the peakdet function from: https://gist.github.com/sixtenbe/1178136#file-peakdetect-py