Optimization efforts are needed in the case of Mega Telco’s customer churn model to get better scores and hence enhance forecasting performance to increase business value. Organizations are attempting to build algorithms that will allow them to predict which clients are most likely to change and take appropriate action (Klepac, 2014; Yu et al., 2018). To identify prospective customer scenarios where churn may be prevented, high accuracy is required, as the client returns are protected, which should offset the expenses of the appropriate retention initiatives.
Successful targeting of churn through a simple ‘yes’ or ‘no’ value basis seems to be an improper model, potentially leading to the data imbalance. The technique of predicting categorical variables via binary classification, in which the output is confined to two classes, has a variety of restrictions and drawbacks (Farhaoui, 2020). Although obvious and simple to understand, it fails to function effectively when the data signal is weak in comparison to the class imbalance signal (Provost and Fawcett, 2013). It has no way of expressing doubt about a specific forecast. A machine learning model’s performance is harmed by an imbalanced class distribution (Santharam and Krishnan, 2020). As a result, upsampling or downsampling should be utilized to resolve the issue. Without resorting to upsampling or downsampling, the error rate can be extremely sensitive to false positives or false negatives (Turkmen et al., 2020). Another potential consideration is entailing cross-validating and verifying various parameters, which increases the process’ accuracy and allows to add more mutually exclusive parameters and data sets (Lalwani et al., 2022).
Additional probabilistic concerns should be taken into account. While the model is more or less suited for churn analysis and targeting, more research will help the structure and churn reduction. In this example, a time series dataset comprising all client information for up to two years would be beneficial for applying time-series domain models like ARIMA models (Lim and Zohren, 2021). This will result in more accurate findings as well as the ability to cross-validate over a larger scene.
Another factor to consider is the study’s limitations, which influenced the outcomes. Although the amount of observations is enough, other datasets, such as customer geographic location and competition information, are required (Lalwani et al., 2022). In this case, the best strategy is to identify a time series dataset encompassing all client data for up to two years in order to achieve the greatest results for forecasting and making market judgments in the future. More information from the result can be extracted within these additional data sets. Because the model is so reliant on the simplicity of interpretation, neural networks or strong gradient boosting can be used to increase accuracy (De Caigny et al., 2020).
Reference List
De Caigny, A., Coussement, K., De Bock, K.W. and Lessmann, S. (2020) ‘Incorporating textual information in customer churn prediction models based on a convolutional neural network’, International Journal of Forecasting, 36(4), pp. 1563-1578.
Farhaoui, Y. ed., 2020. Big Data and Networks Technologies. Springer International Publishing.
Klepac, G. ed., 2014. Developing churn models using data mining techniques and social network analysis. IGI Global.
Lalwani, P., Mishra, M.K., Chadha, J.S. and Sethi, P. (2022) ‘Customer churn prediction system: a machine learning approach’, Computing, 104(2), pp. 271-294.
Lim, B. and Zohren, S. (2021) ‘Time-series forecasting with deep learning: a survey’, Philosophical Transactions of the Royal Society, 379(2194), pp.202-209.
Provost, F. and Fawcett, T., 2013. Data Science for Business: What you need to know about data mining and data-analytic thinking. ” O’Reilly Media, Inc.”.
Santharam, A. and Krishnan, S.B. (2018) ‘Survey on customer churn prediction techniques’, International Research Journal of Engineering and Technology, 5(11), p.13-16.
Turkmen, A., Bahcevan, C.A., Alkhanafseh, Y. and Karabiyik, E. (2020) ‘User behavior analysis and churn prediction in ISP’, New Trends and Issues Proceedings on Advances in Pure and Applied Sciences, 2(12), pp. 57-67.
Yu, R., An, X., Jin, B., Shi, J., Move, O.A. and Liu, Y. (2018) ‘Particle classification optimization-based BP network for telecommunication customer churn prediction’, Neural Computing and Applications, 29(3), pp. 707-720.