Dimensionality and data reduction in telecom churn prediction

Abstract

Purpose -Churn prediction is a very important task for successful customer relationship management. In general, churn prediction can be achieved by many data mining techniques. However, during data mining, dimensionality reduction (or feature selection) and data reduction are the two important data preprocessing steps. In particular, the aims of feature selection and data reduction are to filter out irrelevant features and noisy data samples, respectively. The purpose of this paper, performing these data preprocessing tasks, is to make the mining algorithm produce good quality mining results. Design/methodology/approach -Based on a real telecom customer churn data set, seven different preprocessed data sets based on performing feature selection and data reduction by different priorities are used to train the artificial neural network as the churn prediction model. Findings -The results show that performing data reduction first by self-organizing maps and feature selection second by principal component analysis can allow the prediction model to provide the highest prediction accuracy. In addition, this priority allows the prediction model for more efficient learning since 66 and 62 percent of the original features and data samples are reduced, respectively. Originality/value -The contribution of this paper is to understand the better procedure of performing the two important data preprocessing steps for telecom churn prediction.

Offering a hybrid approach of data mining to predict the customer churn based on bagging and boosting methods

“…PCA is a multivariate statistic technique that determines key features needed to achieve the best description of variance in a set (Lin et al, 2014). Despite other methods that use a subset of features to reduce dimensions of data, PCA combines principle features to create a subset equal or smaller than the primary features.…”

“…In this section, a summary of the related works in the field of churn customer and other fields provided. In Lin et al (2014) a combination of SOM, association rules, and PCA was used as data pre-processing method and then the whole process was combined with neural networks. The results showed higher performance achieved by SOM+PCA +ANN.…”

“…Furthermore, combination of clustering and classifiers technique, as a modern technique, has drawn attention of many authors in machine learning field. In fact, clustering is pre-processing stage to eliminate outlier data and then clustering is used to develop prediction model (Tsai, 2014;Lin et al, 2014;Hsieh, 2005). Studies have shown that performance of combined clustering method is better than the basic classifier.…”

“…A SOM is a type of ANN that is trained using unsupervised learning to produce a low-dimensional, discretized representation of the input space of the training samples, called a map. The main properties of SOM can be stated as Lin et al (2014):…”

RFM approach for telecom insolvency modeling

“…In essence, not without significance is the time required for the proper reaction. Dealing with three variables implies that data retrieval and preparation for scoring the customers can be reduced to a couple of minutes (depends on customer database size and data warehouse performance) which is important matter to assure high operational performance of the model and make it resistant to noise (Lin et al, 2014;Wilson and Martinez, 2000).…”