Distributed Machine Learning for Predicting Customer Behavior While Preserving Privacy

Analyzing customer behavior has great value, that's why researchers until this day are trying to more understand customer behavior, because this enables them to realize what's behind purchase intention, how and what customers want to buy, and what are the factors that affect customer behavior. Big data, artificial intelligence, and machine learning are the three most critical pillars on which the concept of business analytics is based, which leads to ideal customer behavior analyzes and through which customer behavior can be predicted. The benefits of predicting customer behavior are many, such as increasing sales and profits, maintaining existing customers and attracting more. Therefore, the purpose of this study is to analyze the existing methods for predicting customer behavior in addition to identifying the types of federated learning and the suitable type for each scenario. This study provides a comprehensive description of customer behavior prediction techniques. The authors found that there are many models used to predict customer behavior and that there are four main factors affecting customer behavior. The most prominent approaches to maintain data privacy are differential privacy and homomorphic encryption. Moreover, there are three main types of federated learning which are horizontal, vertical and transfer federated learning. As a future study, one of the federated learning approaches will be used to predict customer behavior.