Customer churn prediction model based on hybrid neural networks DOI Creative Commons
Xinyu Liu, Guoen Xia, Xianquan Zhang

и другие.

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Дек. 28, 2024

Abstract In today’s competitive market environment, accurately identifying potential churn customers and taking effective retention measures are crucial for improving customer ensuring the sustainable development of an organization. However, traditional machine learning algorithms single deep models have limitations in extracting complex nonlinear time-series features, resulting unsatisfactory prediction results. To address this problem, study proposes a hybrid neural network-based model, CCP-Net. data preprocessing stage, ADASYN sampling algorithm balances sample sizes churned non-churned to eliminate negative impact imbalance on model performance. feature extraction CCP-Net uses Multi-Head Self-Attention learn global dependencies input sequences, combines with BiLSTM capture long-term sequential data, CNN extract local ultimately generates Experimental results cross-validation Telecom, Bank, Insurance, News datasets show that outperforms comparison all performance metrics. For example, achieves Precision 92.19% Telecom dataset, 91.96% Bank 95.87% Insurance 95.12% which compares other network models, improvement ranges from 1% 3%. These indicate design effectively improves accuracy robustness prediction, enabling it be widely applied different industries, especially financial, telecommunication, media fields, provide more comprehensive management strategies enterprises.

Язык: Английский

Advancements in Natural Language Processing: Implications, Challenges, and Future Directions DOI Creative Commons
Supriyono Supriyono, Aji Prasetya Wibawa,

Suyono Suyono

и другие.

Telematics and Informatics Reports, Год журнала: 2024, Номер 16, С. 100173 - 100173

Опубликована: Ноя. 7, 2024

Язык: Английский

Процитировано

14
Efficient shrinkage temporal convolutional network model for photovoltaic power prediction DOI
Min Wang, Congjun Rao, Xinping Xiao

и другие.

Energy, Год журнала: 2024, Номер 297, С. 131295 - 131295

Опубликована: Апрель 15, 2024

Язык: Английский

Процитировано

9
MNeuralTab: Integrating meta-modeling and neural networks for customer churn prediction in e-commerce DOI Creative Commons

Arif Mohammad Asfe,

Md. Rashadur Rahman, Md. Sabir Hossain

и другие.

Deleted Journal, Год журнала: 2025, Номер 7(6)

Опубликована: Май 30, 2025

Язык: Английский

Процитировано

0
Customer-Centric Decision-Making with XAI and Counterfactual Explanations for Churn Mitigation DOI Creative Commons
Simona‐Vasilica Oprea, Adela Bârã

Journal of theoretical and applied electronic commerce research, Год журнала: 2025, Номер 20(2), С. 129 - 129

Опубликована: Июнь 3, 2025

In this paper, we propose a methodology designed to deliver actionable insights that help businesses retain customers. While Machine Learning (ML) techniques predict whether customer is likely churn, alone not enough. Explainable Artificial Intelligence (XAI) methods, such as SHapley Additive Explanations (SHAP) and Local Interpretable Model-agnostic (LIME), highlight the features influencing prediction, but need strategies prevent churn. Counterfactual (CF) explanations bridge gap by identifying minimal changes in business–customer relationship could shift an outcome from churn retention, offering steps enhance loyalty reduce losses competitors. These might fully align with business constraints; however, alternative scenarios can be developed achieve same objective. Among six classifiers used detect cases, Balanced Random Forest classifier was selected for its superior performance, achieving highest recall score of 0.72. After classification, Diverse ML (DiCEML) through Mixed-Integer Linear Programming (MILP) applied obtain required features, well range permitted itself. We further apply DiCEML uncover potential biases within model, calculating disparate impact some features.

Язык: Английский

Процитировано

0
Application of GWO-attention-ConvLSTM Model in Customer Churn Prediction and Satisfaction Analysis in Customer Relationship Management DOI Creative Commons
Hui Zhang, Weihua Zhang

Heliyon, Год журнала: 2024, Номер 10(17), С. e37229 - e37229

Опубликована: Сен. 1, 2024

Customer Relationship Management (CRM) is vital in modern business, aiding the management and analysis of customer interactions. However, existing methods struggle to capture dynamic complex nature relationships, as traditional approaches fail leverage time series data effectively. To address this, we propose a novel GWO-attention-ConvLSTM model, which offers more effective prediction churn satisfaction. This model utilizes an attention mechanism focus on key information integrates ConvLSTM layer spatiotemporal features, effectively modeling temporal patterns data. We validate our proposed multiple real-world datasets, including BigML Telco Churn dataset, IBM Cell2Cell Orange Telecom dataset. Experimental results demonstrate significant performance improvements compared baseline models across these datasets. For instance, achieves accuracy 95.17%, recall 93.66%, F1 score 92.89%, AUC 95.00%. Similar are validated other In conclusion, makes advancements CRM domain, providing powerful tools for predicting analyzing By addressing limitations leveraging capabilities deep learning, mechanisms, optimization algorithms, paves way improving relationship practices driving business success.

Язык: Английский

Процитировано

2
Q-ensemble learning for customer churn prediction with blockchain-enabled data transparency DOI
Usama Arshad, Gohar F. Khan, Fawaz Khaled Alarfaj

и другие.

Annals of Operations Research, Год журнала: 2024, Номер unknown

Опубликована: Окт. 17, 2024

Язык: Английский

Процитировано

1
A Novel Deep Convolutional Neural Network Algorithm for Equity Price Prediction DOI Open Access

Jesmine Mary Antony,

Natarajan Sundaram

International Research Journal of Multidisciplinary Technovation, Год журнала: 2024, Номер unknown, С. 275 - 291

Опубликована: Ноя. 30, 2024

Predicting stock prices is one of the difficult issues for researchers and investors. The study suggests an equity price prediction based on feature neural network extraction. We expect using technovative forecasting from traditional Machine Learning (ML) models namely Linear Regression (LR), Autoregressive Integrated Moving Averages (ARIMA), advanced Deep (DL) algorithms such as Long Short-Term Memory Recurrent Neural Network (LSTM-RNN) Convolutional Network-Long (CNN-LSTM). select seven features historical data: date, close, open, high, low, volume, change %. study’s novelty accuracy compared to step-by-step backtesting methodology ML DL algorithms. first use CNN extract data consisting items preceding 10 days 100 days. After that extracted LSTM predict price. Finally, used robotic error measure analysis, MAE, RMSE, R2, assess all four models. CNN-LSTM model provides a consistent forecast measures with maximum exactness ranging 0 1, MAE-0.03, RMSE-0.04, R2-0.98. proposed maintained its efficiency throughout process when LR, ARIMA, LSTM-RNN conducts robustness hypothesis check ANOVA test statistic superior predictability accuracy. In addition, this technique gives academics real-world experience analyzing financial time series confident investment ideas

Язык: Английский

Процитировано

0
Intensified customer churn Prediction: Connectivity with weighted Multi-Layer Perceptron and Enhanced Multipath Back Propagation DOI

S. Arockia Panimalar,

A. S. Krishnakumar,

S. Senthil Kumar

и другие.

Expert Systems with Applications, Год журнала: 2024, Номер unknown, С. 125993 - 125993

Опубликована: Дек. 1, 2024

Язык: Английский

Процитировано

0
Customer churn prediction model based on hybrid neural networks DOI Creative Commons
Xinyu Liu, Guoen Xia, Xianquan Zhang

и другие.

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Дек. 28, 2024

Abstract In today’s competitive market environment, accurately identifying potential churn customers and taking effective retention measures are crucial for improving customer ensuring the sustainable development of an organization. However, traditional machine learning algorithms single deep models have limitations in extracting complex nonlinear time-series features, resulting unsatisfactory prediction results. To address this problem, study proposes a hybrid neural network-based model, CCP-Net. data preprocessing stage, ADASYN sampling algorithm balances sample sizes churned non-churned to eliminate negative impact imbalance on model performance. feature extraction CCP-Net uses Multi-Head Self-Attention learn global dependencies input sequences, combines with BiLSTM capture long-term sequential data, CNN extract local ultimately generates Experimental results cross-validation Telecom, Bank, Insurance, News datasets show that outperforms comparison all performance metrics. For example, achieves Precision 92.19% Telecom dataset, 91.96% Bank 95.87% Insurance 95.12% which compares other network models, improvement ranges from 1% 3%. These indicate design effectively improves accuracy robustness prediction, enabling it be widely applied different industries, especially financial, telecommunication, media fields, provide more comprehensive management strategies enterprises.

Язык: Английский

Процитировано

0