A new hybrid algorithm for feature selection and its application to customer recognition. (24th February 2009)

• Record Type: Journal Article
• Authors: Yan, Luo; Changrui, Yu
• Is Part Of: International journal of services operations and informatics. Volume 4:Number 2(2009)
• Page Start: 146
• Page End: 158
• Publication Date: 2009-02-24
• Subjects: feature selection -- customer recognition -- optimisation algorithms -- nested partitions -- customer features -- nested partitions -- simulated annealing; Service industries -- Management -- Periodicals; Service industries -- Management -- Information technology -- Periodicals; 658.5
• Languages: English
• ISSNs: 1741-539X

Abstract

An important problem in customer recognition process is to select the most valid customer features. The authors present a new optimisation algorithm that combines a global optimisation algorithm called the nested partitions algorithm and the simulated annealing method. The resulting hybrid algorithm NP/SA retains the global perspective of the nested partitions algorithm and the local search capabilities of the simulated annealing method. A detailed application of the new algorithm to a customer recognition problem is also presented. The numerical results suggest that the new framework has great computation efficiency and convergence speed and is very efficient for a difficult customer feature selection problem.