Customer analysis using a deep inferarer classifier and a variable-sensitive clustering algorithm optimized by the Cuckoo search method

Article type: Research Article

Authors: Ghavidel, Motahare | Yadollahzadeh-Tabari, Meisam^{; *} | GolsorkhTabariAmiri, Mehdi

Affiliations: Department of Computer Engineering, Islamic Azad University, Babol Branch, Babol, Iran

Correspondence: [*] Corresponding author. Meisam Yadollahzadeh-Tabari. E-mail: m_tabari@baboliau.ac.ir.

Abstract: In this paper, we proposed classification and clustering algorithms that are proper for analyzing customer-related datasets, which are mostly high-dimensional with too many instances. For the clustering purpose, This paper presents a Cuckoo-Search-based Variable Weighting (CSVW) Clustering algorithm to obtain optimal variable weights of high-dimensional data for each cluster. This paper also proposes a deep Inferarer Classifier for categorizing customers using Bi-Directional Long Short-Term Memory (Bi-LSTM) neural network, which uses a Fuzzy Inferential Classifier on its last layer. The Insurance Company (TIC) and InstaCart datasets are utilized for the experiments and performance evaluation. Simulation results reveal that the proposed clustering algorithm generates appropriate Silhouette and Elbow criteria scores in a few cycles of execution in comparison to ordinal clustering algorithms. Also, the proposed classification algorithm with fuzzy soft-max classifier hits the better Classification Criteria in comparison.

Keywords: Customer clustering, Cuckoo optimization, variable-sensitive clustering, deep learning

DOI: 10.3233/JIFS-230675

Journal: Journal of Intelligent & Fuzzy Systems, vol. 46, no. 1, pp. 339-353, 2024