Prediction of the customer behavior is a subject that is considered to be "the holy grail" in the business. Data mining techniques are not a new subject, but the amount of data that can be processed by the modern computers and the global market that the world has become has opened a lot of opportunities. This paper considers a method for proposal of video materials to the customers in a video on demand (VOD) system, but its broader usage covers any closed system in which the user is identified before the purchase and history of previous user actions is available. By usingthe data from previous purchases in the systemand applying the well-known Apriori algorithm, a set of association rules is generated. An algorithm that uses the history of the client for which the recommendation should be made, compares it with the association rules found previously and produces the prediction for the best fit videos that will be recommended to the customer. The method is simulated using WEKA for the association rules and using T-SQL procedures and functions for the prediction algorithm. Real data from an existing and publicly available VOD (T-home's MAX TV) system is used for the simulation. The data is put in a relational MS SQL database.
Data mining, prediction, Apriori algorithm, association rules, video-on-demand, WEKA.