User interest based text classification: a novel feature extraction method

摘要

Text communication is an important means for network and mobile terminal users. Spam text brings great distress to the user and the service provider. Traditional text classification method failed to consider the influence of user interest toward text classification and spam filtering. It brutally analyzed and classified the content of text communication of the user. The paper proposed a user interest feature extraction algorithm access frequency-inverse class frequency (AF-ICF) based on TF-IDF algorithm. It utilizes operate behaviour of the user of the text communication to summarize the degree of a text item that the user concerns. Compared with the existing text classification method, AF-ICF effectively described the interest distribution of the user. It provides an important reference for other text classification algorithms and has very high practical value.