Robust fuzzy clustering for multiple instance regression

作者：

Highlights：

• We propose a novel multiple instance regression (MIR) algorithm that is based on robust clustering.

• The proposed algorithm can identify the optimal number of linear regression models, the primary instances within each bag, and can assign labels at the instance as well as at the bag level.

• The proposed algorithm was evaluated using synthetic data sets of varying difficulty and was shown to outperform 4 existing MIR algorithms.

• The proposed algorithm was applied to remote sensing data to predict the yearly average yield of a crop and to drug activity prediction. We showed that our approach achieves higher accuracy and more consistent results than existing methods.

摘要

•We propose a novel multiple instance regression (MIR) algorithm that is based on robust clustering.•The proposed algorithm can identify the optimal number of linear regression models, the primary instances within each bag, and can assign labels at the instance as well as at the bag level.•The proposed algorithm was evaluated using synthetic data sets of varying difficulty and was shown to outperform 4 existing MIR algorithms.•The proposed algorithm was applied to remote sensing data to predict the yearly average yield of a crop and to drug activity prediction. We showed that our approach achieves higher accuracy and more consistent results than existing methods.

论文关键词：Multiple instance regression,Fuzzy clustering,Possibilistic clustering,Multiple model regression

论文评审过程：Received 1 May 2018, Revised 3 December 2018, Accepted 25 January 2019, Available online 1 February 2019, Version of Record 15 February 2019.

论文官网地址：https://doi.org/10.1016/j.patcog.2019.01.030