Artificial Intelligence Based Customer Churn Prediction Model for Business Markets

• Published in: Computational intelligence and neuroscience 2022-09, Vol.2022, p.1-14
• Main Authors: Faritha Banu, J., Neelakandan, S., Geetha, B.T, Selvalakshmi, V., Umadevi, A., Martinson, Eric Ofori
• Format: Article
• Language: English
• Subjects: Accuracy; Algorithms; Artificial intelligence; Automation; Communications industry; Customer services; Customers; Data mining; Datasets; Efficiency; Internet service providers; Investments; Machine learning; Particle swarm optimization; Prediction models; Profitability; Software; Swarm intelligence; Telecommunications; Telecommunications industry; Telecommunications services industry
• ISSN: 1687-5265; 1687-5273
• DOI: 10.1155/2022/1703696

The introduction of artificial intelligence (AI) and machine learning (ML) technologies in recent years has resulted in improved company performance. Customer churn forecast is a difficult problem in many corporate sectors, particularly the telecommunications industry. Because customer churns have a direct impact on a company's total revenue, telecommunications firms have begun to develop 76 models to reduce churns at an earlier stage. Previous research has revealed that AI and ML models are effective CCP solutions. According to this viewpoint, this study proposes a unique AI-based CCP model for Telecommunication Business Markets (AICCP-TBM). The AICCP-TBM model's purpose is to control the existence of churners and non-churners in the telecom sector. The proposed AICCP-TBM model employs a Chaotic Salp Swarm Optimization-based Feature Selection (CSSO-FS) method for the best feature assortment. In addition, a Fuzzy Rule-based Classifier(FRC) is used to distinguish between client churners and non-churners. A technique known as Quantum Behaved Particle Swarm Optimization (QPSO) is used to pick the membership functions for the FRC model in order to improve the classification performance of the FRC model. The performance of the AICCP-TBM model is validated using a benchmark CCP dataset and the experimental results are reviewed from several angles. In relations of presentation, the imitation consequences demonstrated that the AICCP-TBM model surpassed the most recent state-of-the-art CPP models. The suggested AICCP-TBM method's comparative accuracy was thoroughly tested on the three datasets used. Using datasets 1-3, this technique obtained better levels of accuracy, with the maximum attainable values being 97.25 %, 97.5 % and 94.33 %. The simulation results for the AICCP-TBM model demonstrated improved prediction performance.