الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 148 |  |  |
| | | from | 148 | | |
Abstract
Summary
The convergence of the data science discipline with the existing academic communities of statistics and calculus leads to vigorous discussions, frequently with the two fields embodying the roles of a technologically adept novice and an elderly expert, respectively. Like any developing field, data science introduces innovative ideas and methods for addressing novel inquiries that demand alternative perspectives on established concepts.
The discipline of data science has generated much discussion, enthusiasm, and investment within colleges and universities in recent years. Through the discipline of statistics, data science extends and expands concepts. The most recent papers investigate the critical importance of statistical thinking in data science both in academia and research, as claimed in the directions (National Research Council, 2014).
Conclusion and Recommendations
Telecommunications businesses are becoming increasingly competitive in predicting client churn. Because acquiring new customers is expensive, predicting customer attrition is becoming essential, if not indispensable, as it will allow the implementation of a good management policy that will contribute to customer loyalty as well as a strategic planning and decision-making process in the company. This thesis discusses the feasibility of employing various machine learning methods (both coupled and individual) to predict client attrition. The IBM Sample dataset was used for training and testing in this study.
6.2 Further Future suggestions
Despite its innovative and unique contribution, this study leaves open questions for future investigation. More research is needed to evaluate the role of future work that can be predicted from the presented work includes the introduction of federated learning models with additional data could be evaluated for the telecom- munitions industry’s customer churn problem. The following suggestions are examples of future research:
i. This study can be expanded in the future by utilizing big data analytics. Social network analysis can be used to determine customer satisfaction with telecom services, and these services can subsequently be supplied to reduce churn.
ii. Additional datasets can also be utilized to boost confidence in the results. the models may be used to various industries such as banking, insurance, and airline prediction accuracy can be compared.
iii. The finding’s generalizability should be evaluated. Although this study demonstrated the efficiency of combining ensemble classification algorithms and hybrid resampling approaches in predicting customer attrition, its analysis was limited to these datasets. Future research can evaluate the generalizability of this study’s findings by using the methodologies provided in this study to other datasets.
iv. Only resampling strategies were used in this study to address the class imbalance. Future customer churn prediction research has addressed the class imbalance using different methods such as cost-sensitive classification.
v. Classification methods can be used in future research to do cost-sensitive classification.
vi. Finally, there is no time series data in the dataset used in this investigation. Future study can use algorithms and resampling approaches to implement time-series analysis. Reinforcement learning applied to time-series customer data is also promising.