Abstract:

In a global world, where e-business is rapidly growing, everyone tries to improve his income. Among all the available strategies, trying to predict the customer’s behavior and to anticipate his expectations is a solution. To achieve this goal, one has to design a model of customer’s behavior, usually by mining the tremendous amount of data stored on the company’s computer. In fact, the data mining process implicitly assumes the homogeneity of the context where data are collected. But the cultural background is different when an Egyptian citizen buys with Amazon or when it is a Chinese citizen. We cannot suggest a Chinese citizen to buy a spoon after buying a bowl: stick would be more appropriate. In that context, the classical techniques of machine learning or data mining can be challenged. This is why we suggest investigating the use of analogical learning, which allows highlighting hidden correspondences between diverse contexts. Analogical learning is not new but we provide a set theoretic approach, very suitable for practical implementation. This way to model customer behavior could enhance the customer experience in a business environment.