predicting the attraction and retention of customers in sports pools in isfahan city using a decision tree: presenting a data mining-based model

The purpose of this research is collective learning techniques and a combination of classification algorithms in data mining, evaluated the performance of these algorithms to predict whether customers of isfahan sports pools would drop over a 6-year period. the collected data pertained to 54,000 customers and covered 22 primary characteristics. to evaluate the features, two algorithms—grasshopper optimization algorithm (goa) and simulated annealing (sa)—were used for feature selection, while classification algorithms, specifically decision tree (dt) and k-nearest neighbors (knn), were employed to classify and recognize customer behavior. the results indicated that the combined goa-dt algorithm, with an accuracy of 90.91%, outperformed the sa-dt algorithm, which achieved an accuracy of 87.95%. furthermore, the combined goa-dt algorithm selected only 7 out of the 22 features that were effective in detecting customer churn, whereas the sa-dt algorithm led to the selection of 8 features. in the subsequent step, the knn algorithm was used instead of the dt algorithm. the results demonstrated that the knn algorithm, when combined with goa and sa, ultimately reached an accuracy of 83.86% for the goa algorithm, with the selection of 9 features. based on these findings, it can be concluded that the combined goa-dt algorithm exhibits the best performance, identifying 7 key characteristics: average monthly recharge of the usage account, number of ticket purchases, customer satisfaction level, discounts received, service quality, monthly services, and free training. therefore, it is recommended that managers make decisions based on these identified characteristics to retain customers and estimate customer behavior using such methods.