Ensemble Learning Predicts Banking Customer Demand

By Dr Silpaja Chandrasekar, PhDReviewed by Susha Cheriyedath, M.Sc.Jan 16 2024

In a paper published in the journal PLOS ONE, researchers proposed a customer demand learning model to enhance the market share of large banks. Based on financial datasets, the model optimized the distribution of bank big data channels to overcome customer demand homogenization.

Comparing random forest (RF) and support vector machine (SVM) models, the ensemble learning approach showed superior accuracy, enabling targeted marketing and enhancing the probability of product marketing success. Applying the ensemble learning model resulted in substantial sales growth and increased customer satisfaction, highlighting its effectiveness in improving overall marketing for bank e-commerce services. The study offers valuable insights for bank marketing decision-making and strategy optimization based on customer demand.

Related Work

Previous research has explored diverse applications of ensemble learning, including efforts to maximize diversity and accuracy, dynamic ensemble algorithms for neural networks, and its role in domains like network security and software defect prediction. Studies also investigated domain adaptive ensemble learning, climate funding, and green bond markets.

Ensemble learning found applications in forest fire detection, energy management, and slope stability prediction. Meanwhile, in bank marketing, studies examined online relationship marketing, customer satisfaction, loyalty, and the impact of artificial intelligence and green banking initiatives on marketing capabilities. The shift from traditional to internet marketing models in the era of big data necessitates a deeper understanding of user structures for optimizing algorithm models in bank wealth management product marketing.

Ensemble Learning for Banking

In response to the advancing precision requirements of machine learning tasks facilitated by Internet of Things (IoT) technology, this study introduces a robust learning model structure within the ensemble learning framework to practically complete marketing-related tasks. The ensemble learning model employs a hierarchical control decision-making structure for classification, regression, and preprocessing tasks. Illustrating its architecture in the context of bank marketing ability prediction, the model actively showcases the data interaction mechanism among the user, server, and data interaction layers.

Accurate customer behavior prediction, particularly regarding time deposits, is crucial in banking. The paper details the extraction process of financial data from significant banks using extract, transform, load (ETL) tools and adheres to ethical and privacy standards. Researchers actively analyze the collected financial data to enhance the model's predictive capacity. Outlining the structure of the customer data analysis and management system provides insights into the server-side framework for bank marketing businesses.

The study introduces an ensemble learning-based bank marketing ability forecasting model that combines the strengths of extreme gradient boosting (XGBoost) and RF algorithms. The model employs a comprehensive sampling strategy and parameter adjustments to improve accuracy. The schematic diagram illustrates the integration of RF for feature selection and preliminary predictions, followed by XGBoost for capturing complex relationships and patterns in the data.

The input data is cleaned and pre-processed to evaluate the performance of the ensemble learning model. The experimental setup involves comparative analysis with other machine learning algorithms, including decision trees, RF, SVM, and k-nearest neighbor (KNN). Performance metrics are employed, including accuracy, precision, recall rate, F1 value, data transmission delay, and overall system delay. The findings demonstrate the superior predictive capability of the ensemble learning-based marketing ability model, enabling targeted marketing and relationship strengthening between customers and banks.

Ensemble Model Dominates Bank Predictions

This study actively assesses the ensemble learning model against various models, such as decision trees, RF, SVM, and KNN, while considering a model proposed in 2022. The change in prediction accuracy is showcasing a rapid rise followed by stabilization. Notably, the proposed ensemble learning model achieves a high accuracy of 91%, outperforming other models. Precision, recall rate, and F1 value trends consistently demonstrate the superior performance of the ensemble learning model.

This method depicts a stable accuracy trend after 100 iterations, where the proposed model excels with 91% accuracy. In terms of precision, recall rate, and F1 value, the ensemble learning model consistently outperforms other models, showcasing its effectiveness in predicting bank marketing ability.

Analyzing the data transmission performance of various models reveals fluctuations in data transmission delay with increasing iteration times. The ensemble learning model consistently exhibits lower data transmission delays than decision tree and SVM models. This model also demonstrates the numerical change in overall network delay, where the ensemble learning model maintains a favorable position with lower delays.

It emphasizes the superiority of the ensemble learning model, showcasing a higher sales growth rate (25.67%) and customer satisfaction growth rate (20.52%) compared to alternative algorithms. Even when compared to the model introduced in 2022, the proposed model exhibits better performance in both metrics.

Conclusion

In summary, this paper explores the impact of new marketing strategies in the banking sector, emphasizing the use of ensemble learning for data analysis and prediction. The study highlights the potential for targeted marketing and improved customer-bank relationships by contrasting RF and SVM models. However, there are limitations, such as the need for longer data spans and parameter adjustments. Future research should address these aspects to enhance the model's predictive performance and practical applicability in bank marketing ability prediction.