Towards Ensemble Learning for Tracking Food Insecurity From News Articles

Abstract

The study integrates ensemble learning into a task of classifying if a news article is on food insecurity or not. Similarity algorithms were exploited to imitate human cognition, an innovation to enhance performance. Four out of six classifiers generated performance improvement with the innovation. Articles on food insecurity identified with best classifier were generated into trends which were comparable with official trends. This paper provides information useful to stake holders in taking appropriate action depending on prevailing conditions of food insecurity. Two suggestions are put forth to promote performance: (1) using articles aggregated from several news media and (2) blending more classifiers in an ensemble.