This is the code for the Machine Learning for Quantified Self (ML4QS) in VU Amsterdam MSc Computer Science.
Recently, the sensory data from subjects were widely used in predicting human behavior through using multiple machine learning methods. Considering the quantified self data is still worth to analyze in-depth, we proposed to describe an exquisite pipeline of data preprocessing, feature engineering and feature selection. Then, typical machine learning methods and an ensemble model were evaluated and compared based on the result of prediction.
activity-classification-gbc-fourier-transform.ipynb notebook is a fourier transform tutorial.
ml4qs-benchmark.ipynb notebook contains two powerful existed models, separately GradientBoostingClassifier and lightgbm we used to work as a benchmark.
clustrer&ensemble.ipynb notebook contains feature selection procedure and an ensemble machine learning-based algorithms imported to produce a predictionand compared to the benchmark.