Features weight estimation using a genetic algorithm for customer churn prediction in the telecom sector
Source of Publication
Advances in Intelligent Systems and Computing
© Springer Nature Switzerland AG 2019. The high dimensional dataset results in more noise, will require more computations, has huge sparsity linked with high dimensional features and has thus introduced great challenges in data analysis. To efficiently manipulate and address the impact of the challenges faced by high dimensional dataset, researchers used several features reduction methods. Feature reduction is a formidable step, when dealing with improving the accuracy and reducing the processing time, within high-dimensional data; wherein the feature set is reduced before applying data mining or statistical methods. However with attribute reduction, there is a high chance of loss of important information. In order to avoid information loss, one way is to assign weights to the attributes through domain-expert which is a subjective exercise. It is not only costly but also requires human-expert of the field. Therefore, there is a need for a technique to automatically assign more appropriate weights without involving domain expert. This paper presents a novel features weighting technique. The technique employs a genetic algorithm (GA) to automatically assign weights to the attributes based on Naïve Bayes (NB) classification. Experiments have been conducted on publically available dataset to compare the performance of the proposed approach and NB approach without the weighted features for predicting customer churn in telecommunication sector. The experimental results have demonstrated that the proposed technique outperformed through achieving an overall 89.1% accuracy, 95.65% precision which shows the effectiveness of the proposed technique.
Attribute weighting, Classification, Genetic algorithm, Naïve Bayes
Amin, Adnan; Shah, Babar; Abbas, Ali; Anwar, Sajid; Alfandi, Omar; and Moreira, Fernando, "Features weight estimation using a genetic algorithm for customer churn prediction in the telecom sector" (2019). All Works. 1661.
Indexed in Scopus