The telecommunications industry is experiencing rapid transformation, resulting in tense competition and increased customer volatility. Telecom churn, which refers to the discontinuation of services by customers, poses a serious challenge due to its direct impact on revenue and long-term profitability. Addressing this issue requires effective methods for understanding and predicting customer behavior. Hence, a logic mining approach is introduced in this study, namely the Permutation Weighted Random 2 Satisfiability Reverse Analysis Method, to classify customer churn in the telecommunications sector. The proposed method is based on a logical rule known as Weighted Random 2 Satisfiability, which is implemented in the Discrete Hopfield Neural Network. The logical rule facilitates the dynamic allocation of negative literals, contributing to improved logical representation. Furthermore, the Election algorithm is incorporated during the training phase to enhance the accuracy of logical structure interpretation. The proposed method is capable of extracting optimal data patterns and generating induced logic that accurately describes customer churn behaviour. This induced logic not only predicts whether a customer will churn but also provides interpretable insights into the underlying causes. Experimental results demonstrate a strong average accuracy of 85.6%, indicating the effectiveness and scalability of the proposed approach for knowledge discovery. Although the proposed approach achieves strong accuracy, the lower F1-Score and Matthews Correlation Coefficient reveal limitations in churn customer classification, highlighting the need for further improvement in handling class imbalance. This study contributes to the field of data mining by offering a logic-based framework for churn classification and emphasizing its practical relevance in supporting strategic customer retention efforts in a competitive telecommunications sector.
Zamri et al. (Wed,) studied this question.