Data Mining - Concepts and Techniques.

Data Mining Practical Machine Learning Tools and Techniques

International Journal of Computer and Applications

Imbalanced data learning is one of the challenging problems in data mining; among this matter, founding the right model assessment measures is almost a primary research issue. Skewed class distribution causes a misreading of common evaluation measures as well it lead a biased classification. This article presents a set of alternative for imbalanced data learning assessment, using a combined measures (G-means, likelihood ratios, Discriminant power, F-Measure Balanced Accuracy, Youden index, Matthews correlation coefficient), and graphical performance assessment (ROC curve, Area Under Curve, Partial AUC, Weighted AUC, Cumulative Gains Curve and lift chart, Area Under Lift AUL), that aim to provide a more credible evaluation. We analyze the applications of these measures in churn prediction models evaluation, a well known application of imbalanced data Keywords: imbalanced data, Model assessment, accuracy , G-means, likelihood ratios, F-Measure, Youden index, Matthews correlation coefficient, ROC, AUC, P-AUC,W-AUC, Lift, AUL

/pdf/evaluation-measures-for-models-assessment-over-imbalanced-22i4c1anx8.pdf

Evaluation Measures for Models Assessment over Imbalanced Data Sets

A comparison of machine learning techniques for customer churn prediction

Customer churn in the mobile telephony industry is a continuous problem owing to stiff competition, new technologies, low switching costs, deregulation by governments, among other factors. To address this issue, players in this industry must develop precise and reliable predictive models to identify the possible churners beforehand and then enlist them to intervention programs in a bid to retain as many customers as possible. This paper proposes a new set of features with the aim of improving the recognition rates of possible churners. The features are derived from call details and customer profiles and categorized as contract-related, call pattern description, and call pattern changes description features. The features are evaluated using two probabilistic data mining algorithms Naive Bayes and Bayesian Network, and their results compared to those obtained from using C4.5 decision tree, a widely used algorithm in many classification and prediction tasks. Experimental results show improved prediction rates for all the models used.

/pdf/predicting-customer-churn-in-mobile-telephony-industry-using-3d9mp4gjbx.pdf

Predicting Customer Churn in Mobile Telephony Industry Using Probabilistic Classifiers in Data Mining

Class imbalance is a major problem that is often experienced when dealing with rare events, such as churn recognition in the mobile telecommunications industry. In this work, various strategies of addressing the problem are studied and a demonstration of how under-sampling and Synthetic Minority Oversampling Technique (SMOTE) can be used to address the problem is given. The two techniques are implemented individually first, and then we take the hybrid approach by combining both SMOTE and undersampling. For performance evaluation, two predictive techniques, C4.5 decision tree and Naive Bayes classifier with 10-fold cross validation are used. TPR and FPR values are obtained and used to generate ROC curves from which AUC values are calculated and performance comparison of the three techniques is performed. Results show that the hybrid approach achieves better performance.

/pdf/handling-class-imbalance-in-mobile-telecoms-customer-churn-2vnw0o7zzw.pdf

Li Hong

Papers

Predicting Customer Churn in Mobile Telephony Industry Using Probabilistic Classifiers in Data Mining

Handling Class Imbalance in Mobile Telecoms Customer Churn Prediction