An experimental study on four models of customer churn prediction

Zhu, Chao; Qi, Jiayin; Wang, Chen

doi:10.1109/icsmc.2009.5346170

Cited by 9 publications

(4 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Nowadays, the prediction of the customer churn rate is becoming an increasingly important and complex operation in every organization, hence the need and the obligation to develop new models and very efficient algorithms. 36…”

Section: Related Workmentioning

confidence: 99%

Hybrid approach using machine learning algorithms for customers' churn prediction in the telecommunications industry

Beeharry

Fokone

2021

Concurrency and Computation

View full text Add to dashboard Cite

Ensemble learning involves using several individual classifiers and combining their predictions, which may result in better performance than a single classifier. This article proposes a two-layer flexible voting ensemble model to predict the customer churn rates in the telecommunication industries. The datasets used in this study are from IBM Sample Data Sets and Duke University. Following the pre-processing stage, the datasets have been categorized into an imbalanced and a balanced set. The balanced set comprises of an equal number of instances for both classes ('churn' and 'not churn').Extensive investigations were also carried out to determine the circumstances under which the model provides the best performance. The results of the hybrid algorithm with the IBM imbalanced dataset give an accuracy of 82.30% and an F1-score of 63%. However, with the IBM balanced dataset, an accuracy of 76.20% and an F1-score of 77.06% are obtained. When considering the dataset from Duke University an accuracy of 71.33% and an F1-score of 14.3% are obtained with the imbalanced dataset. The proposed model provides an accuracy of 60.41% and an F1-score of 64.13% with the corresponding balanced dataset. Test work results indicate that the approach adopted has significantly increased the F1-score of the classification when considering a balanced dataset in both cases. Additionally, p-values of less than 0.05 indicate that the results obtained with IBM imbalanced dataset and both balanced and imbalanced dataset from Duke University are statistically significant.

show abstract

Section: Related Workmentioning

confidence: 99%

Hybrid approach using machine learning algorithms for customers' churn prediction in the telecommunications industry

Beeharry

Fokone

2021

Concurrency and Computation

View full text Add to dashboard Cite

show abstract

“…The standard Churn rate is revealed by segment (a) above, but we used similar methodology to obtain churn rates for other segments (see [1,5,[12][13]). We have observed that overall the customers of the CP have a high Churn or attrition rate, although the figures that we obtain using conventional (year on year) analysis may be slightly distorted -although churn rates are often high in SMEs [4].…”

Section: S(d)mentioning

confidence: 99%

Towards Accurate Predictions of Customer Purchasing Patterns

Valero-Fernandez¹,

Collins

Rigby³

et al. 2017

2017 IEEE International Conference on Computer and Information Technology (CIT)

View full text Add to dashboard Cite

A range of algorithms was used to classify online retail customers of a UK company using historical transaction data. The predictive capabilities of the classifiers were assessed using linear regression, Lasso and regression trees. Unlike most related studies, classifications were based upon specific and marketing focused customer behaviours. Prediction accuracy on untrained customers was generally better than 80%. The models implemented (and compared) for classification were: Logistic Regression, Quadratic Discriminant Analysis, Linear SVM, RBF SVM, Gaussian Process, Decision Tree, Random Forest and Multi-layer Perceptron (Neural Network). Postcode data was then used to classify solely on demographics derived from the UK Land Registry and similar public data sources. Prediction accuracy remained better than 60%.

show abstract

“…In contrast to the decision tree, which establishes a probability for a prediction symmetrical with neural networks [8] and Logistic Regression, there are other more categorization methods accessible. The processes involved in predicting customer turnover have become more sophisticated today in every firm, making it necessary to create some fresh and potent models [9]. Thus, churn management, particularly in the applications of the mobile telephony industry, places a high priority on customer churn prediction.…”

Section: Introductionmentioning

confidence: 99%

Bio Inspired Approach for Customer Churn Prediction

Ramesh¹

2022

Preprint

View full text Add to dashboard Cite

Customers are the bottom for any commercial enterprise fulfillment and this is why corporations come to be privy to the importance of obtaining pride of clients. Customer churn exercise is important in aggressive and swiftly growing in telecom sector. Due to superior services, competitive pricing, or other benefits offered by competitor companies to clients who sign up, customers move from one telecom service provider to another. The prediction of customer churn has evolved into an essential component of planning processes and strategic decision making in the telecom sector due to the higher costs associated with gaining new consumers. The prime goal line of the introduced study is to scrutinize exactly how a massive bio-inspired data platform might be used to expect customer’s wear and tear in the application of telecom industry. The probabilities of customers leaving a business have been estimated using Mimosa Pudica optimization (MPO) of two upmost classifying approaches. Big data combined with logistic regression and machine learning is also used in this study for comparisons. Feature selection, reduction and clustering are also performed with the help of MPO to forecast customer revenue in the telecom industry, in order to calculate the probability of churn as a function of a collection of factors or client characteristics. In a similar way, MPO is then used to optimize the two customer's churn classifiers on how close their product in every class. This study predicts and analyses churn using data from two different websites. As a result of these judgments, expectations of consumer churn can efficiently provide an accuracy rate of the best classifier as 0.95 percent and an AUC, area under curve of about 0.96 percent.

show abstract

An experimental study on four models of customer churn prediction

Cited by 9 publications

References 6 publications

Hybrid approach using machine learning algorithms for customers' churn prediction in the telecommunications industry

Hybrid approach using machine learning algorithms for customers' churn prediction in the telecommunications industry

Towards Accurate Predictions of Customer Purchasing Patterns

Bio Inspired Approach for Customer Churn Prediction

Contact Info

Product

Resources

About