Using an Ensemble Learning to Identify the Author of Saikaku’s Posthumous Works

Abstract

In this article, we focus on Saikaku’s posthumous works. Saikaku Ihara (c. 1642～93) is a fiction writer of the Genroku period (1688～1704) in Japan. His researchershave tried to identify his works but problems continue to exist. It remains unclearwhich works were really written by Saikaku especially his posthumous works. This pa-per examines the author of his posthumous works using Random Forests, Boosting andBagging. First, we examined Saikaku and Dansui’s works. A unigram of Japanese par-ticles is the best variable in our research. Among these algorithm, the preferred orderof classification accuracy rate is found to be Random Forests > AdaBoost > Bagging.Then, we examined his posthumous works using the usefulness 9 futures that we gotthe first analysis. It was found that from the result that most chapters are classifiedunder Saikaku, whereas some chapters are classified under Dansui. From what has beendiscussed above, we can conclude that Saikaku’s posthumous works are more likely tobe Saikaku’s than Dansui’s. The result supports the assumption that Dansui relativelyedit the Saikaku’s draft with sufficient accuracy rate however our result also providespotential to has become increasing Dansui’s editing throughout five years.