Most of you know what recommendations by retailers, e-commerce are. Few people, however, know how they are produced behind the web-screens or e-mails. So I would like to explain the mechanism about production of recommendations in a series of my blog as the project is going on in the company. First, I would like to consider three points below one by one. I focus on personalized recommendations, which can be customized to customers individually, based on the information about them. Unpersonalized recommendations, such as recommendations based on just sales ranking are out of our interest and scope because this is expected to be provided for every customer equally.
Of course, customers are the most important for our businesses. The problem is that how customer information is obtained and how customer preference can be known based on customer information. It is clear that the best way to know customer preference is just asking them. But it is almost impossible to ask everything about their preference. Fortunately websites and smart phone are widely used among customers so they make us easier to obtain customer information through “What they view longer”, “What they put into favorite items ” and “What they bought in the past”. We can know customer preference based on this information.
Items mean not only products, but news, information, services and anything which can be chosen by customers. Each item can be expressed by some features, which are characteristics of the products. When two items share the same features, one can be recommended to customers who bought the other because both products has similarities each other. It may be difficult to choose good features to do it. So I would like to continue to research how to choose features effectively.
3. Relationship between customers and items
Once information of customers and items are obtained, the relationship between customers and items should be considered. I imagine it is very important to obtain the relationship accurately, so that recommendations can be accurate and effective. Statistical models are needed to calculate metrics in order to express the strength of the relationship between customers and items.
These three points above are critically important to construct a recommender engine because recommendations are a kind of matching between customers and items. I would like to expand this argument to develop algorithms so that recommendations can be calculated correctly. I found that a dozen of programs of recommender engines, which are open source, are available to us. I would like to review some of them going forward. I hope you can enjoy them, too!