Abstract: The increasing prevalence of harassing calls has seriously affected the normal life of people and social harmony. With the promulgation of the Special Action Plan for Comprehensive Rectification of Harassing Call, a nationwide action against harassing calls has been launched. This paper proposed a harassing call detection framework using mobile operator data. Typical features of harassing calls and profiles of called users are extracted. Outlier detection algorithm is used to mitigate the serious consequences of malicious call marking to a certain extent.And multi-source features are used to train the harassing call detection model, so as to effectively improve the detection efficiency.

Key words: mobile operator data, harassing call detection, multi-source feature