Customer Churn Prediction Research Articles

Smart evaluation for deep learning model: churn prediction as a product case study

Customer churn prediction recently is one of the vital issues that confronts diverse business industries to sustain the customers base and profits. On the other hand, data scientists employ gigantic customer data to automate the data modelling process to offer these models as a generally portable service. This research has two main contributions: deep learning customer churn prediction model and smart evaluation prediction model service. So, this service harnesses any customer data to automate building, evaluation, and deployment of the churn prediction model. The research consists of three main parts. Firstly, it illustrates the dataset labelling which annotates customers data into churn or non-churn. Secondly, the deep learning churn prediction framework using convolutional neural network (CNN) algorithm. Finally, a case study is presented to show how churn prediction service is automatically trained and generated based on real customer data, where CNN parameters are adapted to achieve the most reliable performance in line with customers' behavior. The applied case study achieves accuracy 0.77, area under the curve (AUC) 0.84 and f1 score 0.83.

Data Visualization and Prediction for Telecom Customer Churn

With the deepening of telecom industry reform and the intensification of competition, the customer churn rate of telecom enterprises is gradually increasing. How to predict and effectively reduce customer churn is directly related to the survival and development of telecom enterprises. In order to effectively deal with unbalanced classification and improve the accuracy of high-value customer churn prediction in telecom industry, this paper uses telecom customer data set from kaggle platform to analyze people's use of telecom services, and help telecom operators find out the reasons for customer churn, and establish churn prediction model to reduce customer churn rate. In this paper, firstly, the data set is imported, and then the data visualization analysis is carried out. Then, the random forest model, SVM model and GBDT model are introduced for comparison. Experiments show that random forest has better classification performance than other methods, and improves the accuracy of high-value customer churn prediction.

Predicting customer churn based on changes in their behavior patterns

Customer retention is one of the most important tasks of a business, and it is extremely important to allocate retention resources according to the potential profitability of the customer. Most often the problem of predicting customer churn is solved based on the RFM (Recency, Frequency, Monetary) model. This paper proposes a way to extend the RFM model with estimates of the probability of changes in customer behavior. Based on an analysis of data relating to 33 918 clients of a large Russian retailer for 2019–2020, it is shown that there are recurring patterns of change in their behavior over a single year. Information about these patterns is used to calculate the necessary probability estimates. Incorporating these data into a predictive model based on logistic regression increases prediction accuracy by more than 10% on the metrics AUC and geometric mean. It is also shown that this approach has limitations related to the disruption of behavioral patterns by external shocks, such as the lockdown due to the COVID-19 pandemic in April 2020. The paper also proposes a way to identify these shocks, making it possible to forecast degradation in the predictive ability of the model.

Correction to: Improve customer churn prediction through the proposed PCA‑PSO‑K means algorithm in the communication industry

Correction to: Improve customer churn prediction through the proposed PCA‑PSO‑K means algorithm in the communication industry

Data analysis and machine learning in the context of customer churn prediction

Due to the fierce competition in the market, customers are often faced with multiple choices when choosing products and services. So many industries, including banking, are now facing the problem of how to address customer churn. At the same time, in order to improve the quality of service for users, banks and other institutions need to conduct in-depth research on the characteristics of customers. This paper provides solutions to the above two problems by using data analysis and mining technology and machine learning technology in artificial intelligence. The study provides an in-depth exploration of cus-tomer churn likelihood by analyzing customer behavior and characteristics. This study used data analysis methods such as Chi-square test, Mann Whitney U test, linear regres-sion, and machine learning methods such as logistic regression, random forest, and XGBoost. This research uses public datasets on Kaggle. This study uses data analysis techniques to provide recommendations on how the banking industry should improve service quality, and establishes 6 models with better performance to predict customer churn. In addition, this paper also uses a variety of evaluation indicators to compare the model performance, and selects the random forest model with high predictive ability as the most suitable model. In addition, the order of importance of the factors responsible for customer churn was successfully derived.

Exploiting time-varying RFM measures for customer churn prediction with deep neural networks

AbstractDeep neural network (DNN) architectures such as recurrent neural networks and transformers display outstanding performance in modeling sequential unstructured data. However, little is known about their merit to model customer churn with time-varying data. The paper provides a comprehensive evaluation of the ability of recurrent neural networks and transformers for customer churn prediction (CCP) using time-varying behavioral features in the form of recency, frequency, and monetary value (RFM). RFM variables are the backbone of CCP and, more generally, customer behavior forecasting. We examine alternative strategies for integrating time-varying and non-variant customer features in one network architecture. In this scope, we also assess hybrid approaches that incorporate the outputs of DNNs in conventional CCP models. Using a comprehensive panel data set from a large financial services company, we find recurrent neural networks to outperform transformer architectures when focusing on time-varying RFM features. This finding is confirmed when time-invariant customer features are included, independent of the specific form of feature integration. Finally, we find no statistical evidence that hybrid approaches (based on regularized logistic regression and extreme gradient boosting) improve predictive performance—highlighting that DNNs and especially recurrent neural networks are suitable standalone classifiers for CCP using time-varying RFM measures.

Customer Churn Prediction in the Telecommunication Industry

Customer churn is essential for telecom fields because it reduces income when a customer switches from one service provider to another. It is important for managers to help them find the factors influencing customer churn so that they can make decisions and optimize services. In this essay, SPSS will be used to analyze the factors that influence customer churn. We use contrastive analysis to find out the factors of the customer churn and use logistic regression to analyze the degree of influence of different factors. The result indicates phone charge, quality and diversity of services have impacts on customer churn.

Prediction of Telecom Customer Churn Based on MIPCA-XGBoost Method

In order to solve the problem that the nonlinear information of data in the field of telecom customer churn prediction is not fully used, or even ignored, which leads to inaccurate prediction, this paper introduces the mutual information feature selection method (MIPCA) to filter the features and reduce the dimensions of customer data, and proposes an XGBoost method based on the mutual information feature selection method(MIPCA-XGBoost), which improves the accuracy of the prediction results. By using the data set of telecom industry customers published on Kaggle website, compares the prediction result of this method with that of machine learning algorithms commonly used in this field, and proves the accuracy, recall and F_Score of MIPCA-XGBoost method is higher than other algorithms.

Deep Churn Prediction Method for Telecommunication Industry

Being able to predict the churn rate is the key to success for the telecommunication industry. It is also important for the telecommunication industry to obtain a high profit. Thus, the challenge is to predict the churn percentage of customers with higher accuracy without comprising the profit. In this study, various types of learning strategies are investigated to address this challenge and build a churn predication model. Ensemble learning techniques (Adaboost, random forest (RF), extreme randomized tree (ERT), xgboost (XGB), gradient boosting (GBM), and bagging and stacking), traditional classification techniques (logistic regression (LR), decision tree (DT), and k-nearest neighbor (kNN), and artificial neural network (ANN)), and the deep learning convolutional neural network (CNN) technique have been tested to select the best model for building a customer churn prediction model. The evaluation of the proposed models was conducted using two pubic datasets: Southeast Asian telecom industry, and American telecom market. On both of the datasets, CNN and ANN returned better results than the other techniques. The accuracy obtained on the first dataset using CNN was 99% and using ANN was 98%, and on the second dataset it was 98% and 99%, respectively.

Ensemble Methods in Customer Churn Prediction: A Comparative Analysis of the State-of-the-Art

In the past several single classifiers, homogeneous and heterogeneous ensembles have been proposed to detect the customers who are most likely to churn. Despite the popularity and accuracy of heterogeneous ensembles in various domains, customer churn prediction models have not yet been picked up. Moreover, there are other developments in the performance evaluation and model comparison level that have not been introduced in a systematic way. Therefore, the aim of this study is to perform a large scale benchmark study in customer churn prediction implementing these novel methods. To do so, we benchmark 33 classifiers, including 6 single classifiers, 14 homogeneous, and 13 heterogeneous ensembles across 11 datasets. Our findings indicate that heterogeneous ensembles are consistently ranked higher than homogeneous ensembles and single classifiers. It is observed that a heterogeneous ensemble with simulated annealing classifier selection is ranked the highest in terms of AUC and expected maximum profits. For accuracy, F1 measure and top-decile lift, a heterogenous ensemble optimized by non-negative binomial likelihood, and a stacked heterogeneous ensemble are, respectively, the top ranked classifiers. Our study contributes to the literature by being the first to include such an extensive set of classifiers, performance metrics, and statistical tests in a benchmark study of customer churn.

An adaptive learning approach for customer churn prediction in the telecommunication industry using evolutionary computation and Naïve Bayes

Customer churn is a complex challenge for burgeoning competitive organizations, especially in telecommunication. It refers to customers that swiftly leave a company for a competitor. Acquiring new customers has cost the telecommunication industry more than keeping existing customers. Traditionally, customer churn prediction (CCP) models are applied to aid in analyzing customer behavior and achieving prediction accuracy, which allows the telecommunication industry to target prior retention efforts toward them. However, only accurate CCP based on the available data or already trained supervised model is inadequate for efficient churn prediction, as existing approaches have not been shown or designed to learn with the skill of adaptation to respond quickly to changes in the customer behavior or a decision. Therefore, it is essential to design an approach that easily adapts to learn from new decision scenarios and provides instant insights. This study proposes an adaptive learning approach for this perplexing problem of CCP using the Naïve Bayes classifier with a Genetic Algorithm (subclass of an Evolutionary Algorithm) based feature weighting approach. Further, the performance of the proposed approach is evaluated on publicly available datasets (i.e., BigML Telco churn, IBM Telco, and Cell2Cell) which significantly enhances the prediction performance as compared to the baseline classifier (i.e., Naïve Bayes with default setting, Deep-BP-ANN, CNN, Neural Network, Linear Regression, XGBoost, KNN, Logit Boost, SVM, and PCALB methods) in terms of average precision of 0.97, 0.97, 0.98, a recall rate that stands at 0.84, 0.94, 0.97, and F1-score of 0.89, 0.96, 0.97, an MCC of 0.89, 0.96, 0.97, and accuracy 0.95, 0.97, 0.98 on subject datasets, respectively.

Predicting customer churn using grey wolf optimization‐based support vector machine with principal component analysis

AbstractCustomer churn is a challenging problem that can lead to a loss of organizational assets. Organizations need to predict customer churn successfully in order to get rid of potential damages and gain a competitive advantage. The aim of this study is to provide a churn prediction model by including feature selection and optimization in classification. The study performs principal component analysis (PCA) to select the best features, support vector machine (SVM) to predict customer churn, and grey wolf optimization (GWO) to optimize the parameters of SVM. In other words, this study proposes a novel hybrid model called PCA‐GWO‐SVM to enhance the prediction ability in customer churn. A comparison experiment is carried out, evaluating the proposed model with the other classification algorithms. Experimental results show that the proposed PCA‐GWO‐SVM hybrid model produces higher accuracy, recall, and F1‐score than other machine learning algorithms such as logit, k‐nearest neighbors, naive Bayes, decision tree, and SVM.

MSVM: An Approach for Customer Churn Prediction using modified Support Vector Machine and various Machine Learning Techniques

Consumer attrition is a major issue across the world companies and that is one of their primary worries. Organizations, specifically in the telecommunication sector, are working to build technologies to forecast future customer churn due to the obvious direct impact on profits. It's essential to define the variables that lead to customer churn before making the required efforts to minimize churn. An important impact was the creation of an attrition estimation method that aids telecommunication corporations in predicting which consumers are willing to churn. The methodology proposed in this study employs mathematical methodologies on a big data framework to provide a unique strategy to feature design and evaluation. This research work has proposed a customer churn prediction using a modified Support Vector Machine Learning (mSVM) classifier. The significant contribution of this research is that we have changed the distance function of SVM in both training and testing. Similar machine learning algorithms are also validated on similar datasets such as Naïve Bayes (NB), K-nearest Neighbor (KNN) and Decision Tree (DT) with J48 classifiers. The BigML dataset has been used for detecting churn in the telecom industry in a real-time scenario. In an extensive experimental analysis, we demonstrate mSVM which produces higher accuracy over the traditional machine learning algorithm for different cross-validation methods.

Research on Application of Stacking Technique in Telecom Churn Prediction

In today's saturated telecommunication market, operators need to fully tap the value of data, strengthen the management of existing users, and reduce customer churns. In order to further to reduce the maintenance cost of enterprises, it is necessary to apply more efficient data mining technology to the prediction of telecommunication user churns in order to improve the operating profit of enterprises. This paper took a telecommunication customer churn data set on Kaggle as the research data, and explores a high accuracy customer churn prediction model. Logistic Regression was exploited to create customer churn prediction models, and the parameters of each model were optimized. On this basis, the Stacking technique was used to fuse the models, and the best combination method is obtained. After research, this paper mainly drew the following conclusions: First, in a single model, the model built with XG Boost algorithm has the highest prediction accuracy, which is 86.20%. Secondly, combining the accuracy rate of training set, cross-validation set and test set, it can be found that XG Boost and Random Forest have the best performance under Stacking fusion algorithm. The accuracy of the above three datasets is 97.61%, 85.84%, 86.51%, respectively, which improves the prediction accuracy of the prediction model.

CHURN ANALISIS PADA DATA PELANGGAN TELEKOMUNIKASI MENGGUNAKAN ENSEMBLE LEARNING

Intense competition in broadband services will create high opportunities for consumers to switch providers, such as conditions that arise in competition for SMS, telephone, and internet services. The churn rate is the percentage of consumers who stop subscribing to the service. Ideally, this churn percentage is only 5% – 10%, and if it exceeds this figure, it indicates the company's inability to retain customers. A high churn rate indicates a decline in the cellular operator's market share and affects the company's revenue. Based on these problems, it is necessary to analyze the churn behavior of broadband subscribers to determine the dissatisfaction factors of cellular telecommunications consumers. Then predictions are made for customers who tend to churn from provider companies and determine the characteristics of churn and stay customers. The ensemble method is used to detect churn, which consists of several methods, including random forest, boosting, and super learner. Random Forest is proven to produce the best classification method with an excellent ability to predict customer churn, which is 80.1%, with an average usage time of 3 years.

Study of Competent Machine Learning Classifiers for Customer Churn Prediction in Telecommunication Sector

Study of Competent Machine Learning Classifiers for Customer Churn Prediction in Telecommunication Sector

Prediction and Big Data Impact Analysis of Telecom Churn by Backpropagation Neural Network Algorithm from the Perspective of Business Model.

This study aims to transform the existing telecom operators from traditional Internet operators to digital-driven services, and improve the overall competitiveness of telecom enterprises. Data mining is applied to telecom user classification to process the existing telecom user data through data integration, cleaning, standardization, and transformation. Although the existing algorithms ensure the accuracy of the algorithm on the telecom user analysis platform under big data, they do not solve the limitations of single machine computing and cannot effectively improve the training efficiency of the model. To solve this problem, this article establishes a telecom customer churn prediction model with the help of backpropagation neural network (BPNN) algorithm, and deploys the MapReduce programming framework on Hadoop platform. Using the data of a telecom company, this article analyzes the loss of telecom customers in the big data environment. The research shows that the accuracy of telecom customer churn prediction model in BPNN is 82.12%. After deploying large data sets, the learning and training time of the model is greatly shortened. When the number of nodes is 8, the acceleration ratio of the model remains at 60 seconds. Under big data, the telecom user analysis platform not only ensures the accuracy of the algorithm, but also solves the limitations of single machine computing and effectively improves the training efficiency of the model. Compared with that of the existing research, the accuracy of the model is improved by 25.36%, and the running time is shortened by about twice. This business model based on BPNN algorithm has obvious advantages in processing more data sets, and has great reference value for the digital-driven business model transformation of the telecommunications industry.

Correction to: Improve customer churn prediction through the proposed PCA‑PSO‑K means algorithm in the communication industry

Correction to: Improve customer churn prediction through the proposed PCA‑PSO‑K means algorithm in the communication industry

Arithmetic Optimization with Deep Learning Enabled Churn Prediction Model for Telecommunication Industries

Customer retention is one of the challenging issues in different business sectors, and various firms utilize customer churn prediction (CCP) process to retain existing customers. Because of the direct impact on the company revenues, particularly in the telecommunication sector, firms are needed to design effective CCP models. The recent advances in machine learning (ML) and deep learning (DL) models enable researchers to introduce accurate CCP models in the telecommunication sector. CCP can be considered as a classification problem, which aims to classify the customer into churners and non-churners. With this motivation, this article focuses on designing an arithmetic optimization algorithm (AOA) with stacked bidirectional long short-term memory (SBLSTM) model for CCP. The proposed AOA-SBLSTM model intends to proficiently forecast the occurrence of CC in the telecommunication industry. Initially, the AOA-SBLSTM model performs pre-processing to transform the original data into a useful format. Besides, the SBLSTM model is employed to categorize data into churners and non-churners. To improve the CCP outcomes of the SBLSTM model, an optimal hyperparameter tuning process using AOA is developed. A widespread simulation analysis of the AOA-SBLSTM model is tested using a benchmark dataset with 3333 samples and 21 features. The experimental outcomes reported the promising performance of the AOA-SBLSTM model over the recent approaches.

A Novel Telecom Customer Churn Analysis System Based on RFM Model and Feature Importance Ranking

Aim/Purpose: In this paper, we present an RFM model-based telecom customer churn system for better predicting and analyzing customer churn. Background: In the highly competitive telecom industry, customer churn is an important research topic in customer relationship management (CRM) for telecom companies that want to improve customer retention. Many researchers focus on a telecom customer churn analysis system to find out the customer churn factors for improving prediction accuracy. Methodology: The telecom customer churn analysis system consists of three main parts: customer segmentation, churn prediction, and churn factor identification. To segment the original dataset, we use the RFM model and K-means algorithm with an elbow method. We then use RFM-based feature construction for customer churn prediction, and the XGBoost algorithm with SHAP method to obtain a feature importance ranking. We chose an open-source customer churn dataset that contains 7,043 instances and 21 features. Contribution: We present a novel system for churn analysis in telecom companies, which encompasses customer churn prediction, customer segmentation, and churn factor analysis to enhance business strategies and services. In this system, we leverage customer segmentation techniques for feature construction, which enables the new features to improve the model performance significantly. Our experiments demonstrate that the proposed system outperforms current advanced customer churn prediction methods in the same dataset, with a higher prediction accuracy. The results further demonstrate that this churn analysis system can help telecom companies mine customer value from the features in a dataset, identify the primary factors contributing to customer churn, and propose suitable solution strategies. Findings: Simulation results show that the K-means algorithm gets better results when the original dataset is divided into four groups, so the K value is selected as 4. The XGBoost algorithm achieves 79.3% and 81.05% accuracy on the original dataset and new data with RFM, respectively. Additionally, each cluster has a unique feature importance ranking, allowing for specialized strategies to be provided to each cluster. Overall, our system can help telecom companies implement effective CRM and marketing strategies to reduce customer churn. Recommendations for Practitioners: More accurate churn prediction reduces misjudgment of customer churn. The acquisition of customer churn factors makes the company more convenient to analyze the reasons for churn and formulate relevant conservation strategies. Recommendation for Researchers: The research achieves 81.05% accuracy for customer churn prediction with the Xgboost and RFM algorithms. We believe that more enhancements algorithms can be attempted for data preprocessing for better prediction. Impact on Society: This study proposes a more accurate and competitive customer churn system to help telecom companies conserve the local markets and reduce capital outflows. Future Research: The research is also applicable to other fields, such as education, banking, and so forth. We will make more new attempts based on this system.