Title: Artificial fish swarm algorithm-based multilayer perceptron model for customer churn prediction in IoT with cloud environment

Authors: S. Venkatesh; M. Jeyakarthic

Addresses: Government Arts and Science College, Perumbakkam, Chennai, Tamil Nadu 600131, India ' Tamil Virtual Academy, Chennai, Tamil Nadu 600025, India

Abstract: This paper develops a new optimal feature selection and classification-based CCP model in IoT and cloud environment. The proposed model involves four main processes namely IoT-based data acquisition, preprocessing, feature selection (FS) and classification. Here, the hill climbing (HC) with social spider optimisation (SSO) algorithm is applied as a feature selection, where HC is incorporated into the SSO algorithm to improve the convergence rate and local searching capability. Subsequently, the feature reduced data undergo classification by the use of artificial fish swarm algorithm (AFSA) tuned multilayer perceptron (MLP) called MLP-AFSA. The proposed model additionally involves an alarming process to alert the organisation when higher churn rate is attained. The presented MLP-AFSA-based CCP model has reached a maximum predictive accuracy of 93.52%, which is further increased to 94.93% by the inclusion of HC-SSO-based feature selection process.

Keywords: customer churn; cloud; IoT; machine learning; feature selection; parameter tuning.

DOI: 10.1504/IJBIS.2023.134958

International Journal of Business Information Systems, 2023 Vol.44 No.3, pp.442 - 465

Received: 13 Aug 2020
Accepted: 07 Dec 2020
Published online: 22 Nov 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article