Incremental personalized E-mail spam filter using novel TFDCR feature selection with dynamic feature update

作者：

Highlights：

• Novel feature selection function is proposed for personalized e-mail spam filter.

• An incremental learning mechanism is applied to update classifier dynamically.

• A heuristic function is developed for updating feature set to handle concept drift.

• Proposed approach substantially improves the classification accuracy.

• Very low FPR is achieved in dynamically updated incremental filter.

摘要

•Novel feature selection function is proposed for personalized e-mail spam filter.•An incremental learning mechanism is applied to update classifier dynamically.•A heuristic function is developed for updating feature set to handle concept drift.•Proposed approach substantially improves the classification accuracy.•Very low FPR is achieved in dynamically updated incremental filter.

论文关键词：Incremental learning,Spam filter,Feature selection,Support vector machine

论文评审过程：Received 4 September 2017, Revised 1 July 2018, Accepted 21 July 2018, Available online 24 July 2018, Version of Record 16 August 2018.

论文官网地址：https://doi.org/10.1016/j.eswa.2018.07.049