ML/AI experiments using tensorflow
The original WISDM dataset can be downloaded from http://www.cis.fordham.edu/wisdm/dataset.php
Included in the data/ directory, is the transformed dataset. Every 200 sequential original readings are massaged into a single record of 46 features, which are described in the WISDM_ar_v1.1_trans_about.txt.
The transformed dataset is in arff format: data are in comma-separated, lines beginning for '@' describe attributes and sections. Missing features in a sample are represented by '?'. As such, this arff file needs cleaning during the import phase or offline.
The code here plays with different predictors:
- random forest via tensorflow
To be tried:
- scikit's Random Forest regressor