Trains on categorical and date time features of the data and predicts an anomaly score for new data
Categorical Outlier is tool to detect anomalous observations in categorical and datetime features. Most of the techniques that we already have focusses mostly on numeric features. There is no library available which can detect a outlier within categories. This package builds a profile of the categorical using the past observations and gives an outlier score to a new observation on the basis of this profile. A scenario where this library can be very useful will be a predicting unusual driving behaviour. A driver who drives the same route(s) to drive to may be office will show an anomalous behaviour if he takes altogether different route on particular day. He will get a high outlier for thie behaviour. On the contrary, an uber driver drives to new location everytime and hence, a new destination will not be an anomalous behaviour and hence will get a low score. The package also takes combination of categorical features as input.
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size||File type||Python version||Upload date||Hashes|
|Filename, size categoricaloutlier-0.6.1.tar.gz (3.2 kB)||File type Source||Python version None||Upload date||Hashes View|