A New Approach of Shannon Entropy in Recommender Systems

J. López Herrera (Spain)


Recommender Systems, Shannon Entropy, qualitative reasoning, knowledge representation.


In this article, the procedure described is to use the methodology to evaluate the degree of satisfaction of a Product or e-Service and how to use this information in recommender Systems[1]. In order to carry out the recommendation of a new service or product, the purchases of the users and their opinions on similar services are analyzed. Each service is defined by means of a series of attributes whose values are provided by the users by means of satisfaction surveys. We understand service as any type of product or service susceptible to purchase or hiring (supermarket products, electricity, telephone, restaurants, movies, etc). This methodology, by means of the application of the Shannon entropy, permits us to identify, in a qualitative way [2], the preferences of the users without the need for a pre established model. An important characteristic of the method is that the results respect the allowed values in case of qualitative variables with discrete values. We presented an application in the area of e-bussiness, whose objective is to recommend a new product (films in this case) to the portfolio of users. In the test made, an average percentage of recommendation of 90 % on “bought products” is obtained.

Important Links:

Go Back