This the source code for the random forest algorithm developed for the Ubiety forecasting service

Depenencies

• numpy
• sklearn
• pickle
• random

Data set generation

The actual provided dataset has been artificially generated using different probabilities distributions in order to approximate real life expectations for every single feature.

Use the package manager pip to install the dependencies.

Run

python dataGen.py -s 

or

python dataGen.py -a

for generating a new dataset with a single event type or multiple event types respectively. Consider also changing the number of sample to be generated. Default is 5000

Training

Run

python Exist_RF_Algorithm.py -n 20 -s

for training the random forest algorithm with cross validation. -n specifies the number of folds (n=20) on the dataset and its default value is 10. Set the flag -s to save the trained tree as a file for later predictions

Note: The best tree when cross validating is selected and returned or saved by the algorithm

Prediction

Run

python predict.py 

in order to predict the show up probability of a user. A default user is set by default. Consider changing the features for further prediction. A batch of users can also be predicted.

Licenese

MIT