utils.py

class ucas_dm.utils.Evaluator(data_set)

Bases: object

This class provide some methods to evaluate the performance of a recommend algorithm. Two measures are supported for now: Recall and Precision

__init__(data_set)

Parameters:data_set – User view log.

static _log_to_file(algo_dict, **kwargs)

This func will save algorithm’s dict data and it’s performance data to ./performance.log in json format.

Parameters:

• algo_dict – algorithm’s dict format.
• kwargs – Performance data of the algorithm.

evaluate(algo=<ucas_dm.prediction_algorithms.base_algo.BaseAlgo object>, k=[], n_jobs=1, split_date='2000-1-1', debug=False, verbose=False, auto_log=False)

Parameters:

• algo – recommend algorithm
• k – list of integers represent the number of recommended items.
• n_jobs – The maximum number of evaluating in parallel. Use multi-thread to speed up the evaluating.
• split_date – on which date we split the log data into train and test.
• debug – if true, the evaluator will use 5000 instances in data set to run the test.
• verbose – whether to print the total time that evaluation cost.
• auto_log – if true, Evaluator will automatically save performance data to ‘./performance.log’

Returns:

average recall and precision