Integrating cluster analysis with granular computing for imbalanced data classification problem – A case study on prostate cancer prognosis. (November 2018)

• Record Type: Journal Article
• Authors: Kuo, R.J.; Su, P.Y.; Zulvia, Ferani E.; Lin, C.C.
• Is Part Of: Computers & industrial engineering. Volume 125(2018)
• Page Start: 319
• Page End: 332
• Publication Date: 2018-11
• Subjects: Prognosis -- Prostate cancer -- Granular computing -- Class imbalance -- Classification; Engineering -- Data processing -- Periodicals; Industrial engineering -- Periodicals; 620.00285
• DOI: 10.1016/j.cie.2018.08.031
• Languages: English
• ISSNs: 0360-8352

Abstract

Highlights: Deal with the class imbalance problem using the concept of information granulation. Propose three clustering techniques to implement to construct information granules. A case study on prostate cancer prognosis is used to assess the proposed methods. The results can provide better information for the patients' survival conditions. Abstract: Analyzing imbalanced dataset is a critical and challenging task in data mining, since it requires special treatment for clusters with different sizes. Imbalance dataset commonly exists in some domains like medical problems. This study intends to propose a classification algorithm based on information granulation (IG) concept for handling imbalanced dataset. The proposed algorithm assembles data from majority classes into granules to balance the class ratio within the data. The proposed algorithm works in two stages. First stage generates a set of IGs using metaheuristics approaches which is a kind of automatic clustering algorithm including dynamic clustering using particle swarm optimization (DCPSO), genetic algorithm K -means (GA K -means), and artificial bee colony K -means (ABC K -means). The next stage applies classification algorithm to classify the data. In this study, the proposed algorithm is verified using both balance and imbalanced benchmark datasets. Simulation results show that the proposed algorithms have promising classification results. Furthermore, this study also applies the proposed algorithms to prostate cancer prognosis classification problem. The algorithm is employed to predict survival rate of prostate cancer patients based on some medical data. The result shows that the proposed algorithms have lower error rate.