A Two-Stage Machine learning approach for temporally-robust text classification

作者：

Highlights：

• Proposal of an automatic procedure to learn the Temporal Weighting Functions (TWFs) from the training data. based on a cheap strategy to learn the distribution that better suits each domain.

• Proposal of three lazy strategies to incorporate TWFs into traditional ADC algorithms.

• Further evaluation of the proposed strategies using three real-world textual datasets with distinct temporal characteristics.

摘要

•Proposal of an automatic procedure to learn the Temporal Weighting Functions (TWFs) from the training data. based on a cheap strategy to learn the distribution that better suits each domain.•Proposal of three lazy strategies to incorporate TWFs into traditional ADC algorithms.•Further evaluation of the proposed strategies using three real-world textual datasets with distinct temporal characteristics.

论文关键词：Automatic document classification,Temporal weighting function,Fully-Automated machine learning process

论文评审过程：Received 23 March 2017, Revised 20 April 2017, Accepted 21 April 2017, Available online 27 April 2017, Version of Record 8 May 2017.

论文官网地址：https://doi.org/10.1016/j.is.2017.04.004