Customer relationship management using machine learning and statistical approaches

点击翻译按钮获取中文摘要

It is widely recognized nowadays that Customer Relationship Management (CRM) represents one vital**business function that generates long term profit by developing harmonious relationship with customers. The**technological advancement has enabled new approaches - notably machine learning - to be applied for finding**the best CRM strategies. However, in real life applications, the gathering of customer information for analysis**is a non-standardized process and the quality of the data collected cannot be guaranteed. This project aims at**finding a CRM model capable of addressing an amalgamation of these issues while still having a reasonable**degree of generality across the entire population. In order to achieve this objective, different statistical and**machine learning technologies and methodologies will be implemented including logistic regression, decision**trees, neural networks, Bayesian classification, random forest, ensemble approach, support vector machine and**etc. to predict the propensity of customer defection (churn). Data will be pre-processed including missing data**handling, feature selection, and normalization before the learning algorithms are applied in order to improve**the prediction accuracy. The optimal strategy and the comparison of different state-of-the-art methodologies**will be presented. It is expected that the proposed research will bring significant contributions to both**prediction theory and real life applications. Canadian business owners can make use of the research results to**precisely predict customer behaviors in order to increase businesses' profits.