Adnan Amin

Customer churn prediction in the telecommunication sector using a rough set approach

Customer churn is a critical and challenging problem affecting business and industry, in particular, the rapidly growing, highly competitive telecommunication sector. It is of substantial interest to both academic researchers and industrial practitioners, interested in forecasting the behavior of customers in order to differentiate the churn from non-churn customers. The primary motivation is the dire need of businesses to retain existing customers, coupled with the high cost associated with acquiring new ones. A review of the field has revealed a lack of efficient, rule-based Customer Churn Prediction (CCP) approaches in the telecommunication sector. This study proposes an intelligent rule-based decision-making technique, based on rough set theory (RST), to extract important decision rules related to customer churn and non-churn. The proposed approach effectively performs classification of churn from non-churn customers, along with prediction of those customers who will churn or may possibly churn in the near future. Extensive simulation experiments are carried out to evaluate the performance of our proposed RST based CCP approach using four rule-generation mechanisms, namely, the Exhaustive Algorithm (EA), Genetic Algorithm (GA), Covering Algorithm (CA) and the LEM2 algorithm (LA). Empirical results show that RST based on GA is the most efficient technique for extracting implicit knowledge in the form of decision rules from the publicly available, benchmark telecom dataset. Further, comparative results demonstrate that our proposed approach offers a globally optimal solution for CCP in the telecom sector, when benchmarked against several state-of-the-art methods. Finally, we show how attribute-level analysis can pave the way for developing a successful customer retention policy that could form an indispensable part of strategic decision making and planning process in the telecom sector.

Churn Prediction in Telecommunication Industry Using Rough Set Approach

The Customer churn is a crucial activity in rapidly growing and mature competitive telecommunication sector and is one of the greatest importance for a project manager. Due to the high cost of acquiring new customers, customer churn prediction has emerged as an indispensable part of telecom sectors’ strategic decision making and planning process. It is important to forecast customer churn behavior in order to retain those customers that will churn or possible may churn. This study is another attempt which makes use of rough set theory, a rule-based decision making technique, to extract rules for churn prediction. Experiments were performed to explore the performance of four different algorithms (Exhaustive, Genetic, Covering, and LEM2). It is observed that rough set classification based on genetic algorithm, rules generation yields most suitable performance out of the four rules generation algorithms. Moreover, by applying the proposed technique on publicly available dataset, the results show that the proposed technique can fully predict all those customers that will churn or possibly may churn and also provides useful information to strategic decision makers as well.

Customer Churn Prediction in Telecommunication Industry: With and without Counter-Example

The customer churn is a crucial activity in the competitive and rapidly growing telecommunication industry. Due to the high cost of acquiring a new customer, customer churn prediction is one of the greatest importance for project managers. It is important to forecast customer churn behavior in order to retain those customers that will churn or possibly may churn. This study is another attempt which makes use of rough set theory as one-class classifier and multi-class classifier to reveal the trade-off in the selection of an effective classification model for customer churn prediction. Experiments were performed to explore the performance of four different rule generation algorithms (i.e. Exhaustive, genetic, covering and LEM2). It is observed that rough set as one-class classifier and multi-class classifier based on genetic algorithm yields more suitable performance out of four rule generation algorithms. Furthermore, by applying the proposed techniques (i.e. Rough sets as one-class and multi-class classifiers) on publicly available dataset, the results show that rough set as a multi-class classifier provides more accurate results for binary/multi-class classification problems.

A Prudent Based Approach for Customer Churn Prediction

This study contributes to formalize a three phase customer churn prediction technique. In the first phase, a supervised feature selection procedure is adopted to select the most relevant subset of features by laying-off the redundancy and increasing the relevance that leads to reduced and highly correlated features set. In the second phase, a knowledge based system (KBS) is built through Ripple Down Rule (RDR) learner which acquires knowledge about seen customer churn behavior and handles the problem of brittle in churn KBS through prudence analysis that will issue a prompt to the decision maker whenever a case is beyond the maintained knowledge in the knowledge database. In the final phase, a technique for Simulated Expert (SE) is proposed to evaluate the Knowledge Acquisition (KA) in KB system. Moreover, by applying the proposed approach on publicly available dataset, the results show that the proposed approach can be a worthy alternate for churn prediction in telecommunication industry.

Site selection for food distribution using rough set approach and TOPSIS method

Suitable site selection for a specific purpose is a crucial activity, and of the greatest importance to a project manager. Several methods have been proposed by the research community for effective site selection, but all proposed methods incur high costs. This study explores the combination of a rough set theory approach (RSTA) with Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) for suitable site selection for food distribution. This method provides a set of rules to determine different sites, which ultimately can help management develop strategies for suitable site selection. A set of rules for suitable site selection are derived from information related to a practical case, Pakistan Red Crescent Society (PRCS), to demonstrate the prediction ability of RSTA. The results clearly demonstrate that the RSTA model can be a valuable tool for site identification. Rough set theory also assists management in making appropriate decisions based on their objectives while avoiding unnecessary costs. However, while RSTA provides rules to determine the best sites for food distribution, it does not pinpoint the best sites for food distribution. To be more precise and accurate, this work is extended to another multi-criteria decision-making technique solution: the TOPSIS method. By using this method, this study provides the best top priority site for food distribution of PRCS.

A Comparison of Two Oversampling Techniques (SMOTE vs MTDF) for Handling Class Imbalance Problem: A Case Study of Customer Churn Prediction

Predicting the behavior of customer is at great importance for a project manager. Data driven industries such as telecommunication industries have advantage of various data mining techniques to extract meaningful information regarding customer’s future behavior. However, the prediction accuracy of these data mining techniques is significantly affected if the real world data is highly imbalanced. In this study, we investigate and compare the predictive performance of two well-known oversampling techniques Synthetic Minority Oversampling Technique (SMOT) and Megatrend Diffusion Function (MTDF) and four different rule generation algorithms (Exhaustive, Genetic, Covering, and LEM2) based on rough set classification using publicly available data sets. As useful feature extraction can play a vital role not only in improving the classification performance, but also to reduce the computational cost and complexity by eliminating unnecessary features from the dataset. Minimum Redundancy Maximum Relevance (mRMR) technique has been used in the proposed study for feature extraction which not only selects the best feature subset but also reduces the features space. The results clearly demonstrate the predictive performance of both oversampling techniques and rules generation algorithms that will help the decision makers/researcher to select the ultimate one.

A prudent based approach for compromised user credentials detection

Compromised user credential (CUC) is an activity in which someone, such as a thief, cyber-criminal or attacker gains access to your login credentials for the purpose of theft, fraud, or business disruption. It has become an alarming issue for various organizations. It is not only crucial for information technology (IT) oriented institutions using database management systems (DBMSs) but is also critical for competitive and sensitive organization where faulty data is more difficult to clean up. Various well-known risk mitigation techniques have been developed, such as authentication, authorization, and fraud detection. However, none of these methods are capable of efficiently detecting compromised legitimate users’ credentials. This is because cyber-criminals can gain access to legitimate users’ accounts based on trusted relationships with the account owner. This study focuses on handling CUC on time to avoid larger-scale damage incurred by the cyber-criminals. The proposed approach can efficiently detect CUC in a live database by analyzing and comparing the user’s current and past operational behavior. This novel approach is built by a combination of prudent analysis, ripple down rules and simulated experts. The experiments are carried out on collected data over 6 months from sensitive live DBMS. The results explore the performance of the proposed approach that it can efficiently detect CUC with 97% overall accuracy and 2.013% overall error rate. Moreover, it also provides useful information about compromised users’ activities for decision or policy makers as to which user is more critical and requires more consideration as compared to less crucial user based prevalence value.

Cross-company customer churn prediction in telecommunication: A comparison of data transformation methods

Abstract Cross-Company Churn Prediction (CCCP) is a domain of research where one company (target) is lacking enough data and can use data from another company (source) to predict customer churn successfully. To support CCCP, the cross-company data is usually transformed to a set of similar normal distribution of target company data prior to building a CCCP model. However, it is still unclear which data transformation method is most effective in CCCP. Also, the impact of data transformation methods on CCCP model performance using different classifiers have not been comprehensively explored in the telecommunication sector. In this study, we devised a model for CCCP using data transformation methods (i.e., log, z-score, rank and box-cox) and presented not only an extensive comparison to validate the impact of these transformation methods in CCCP, but also evaluated the performance of underlying baseline classifiers (i.e., Naive Bayes (NB), K-Nearest Neighbour (KNN), Gradient Boosted Tree (GBT), Single Rule Induction (SRI) and Deep learner Neural net (DP)) for customer churn prediction in telecommunication sector using the above mentioned data transformation methods. We performed experiments on publicly available datasets related to the telecommunication sector. The results demonstrated that most of the data transformation methods (e.g., log, rank, and box-cox) improve the performance of CCCP significantly. However, the Z-Score data transformation method could not achieve better results as compared to the rest of the data transformation methods in this study. Moreover, it is also investigated that the CCCP model based on NB outperform on transformed data and DP, KNN and GBT performed on the average, while SRI classifier did not show significant results in term of the commonly used evaluation measures (i.e., probability of detection, probability of false alarm, area under the curve and g-mean).