Research And Application Of Attribute Inference Technique Based On Decision Tree Method

Personalized recommendation system is mainly used in the E-commerce. The systemis based on the relationship between the product and the consumer. Furthermore, wewould like to use the relationship between the product and the consumer to makepersonalized recommendations to our customers. Personalized recommendation system isdeveloped based on the application of association rules, collaborative filtering technology,and content-based analysis technology. All three methods above share the same problems,that is, data sparsity problem, real-time problem, and cold start problem. I develop apersonal recommendation method based on deduced consumer’s properties using thedecision tree algorithm.Current personalized recommendation algorithms focus on consumer choice of goods.In a reverse thinking, I use consumers’purchase records to deduce consumers’ properties,and furthermore, making general rules about individual purchasing behavior and usingthem to make personalized recommendations. The key step in this is the deduction ofconsumer properties. I employ the concept of decision tree algorithm to complete thisstep.First, I implement the decision tree algorithm to deduct consumer properties. Itshows in our test runs that it has correct rate of69.32%. Then, we run threerecommendation methods and the results in following show that the recommendationbased on our deduced consumer properties is quite promising.(1) Personal recommendation made by using original data has correct rate45.87%.(2) Person recommendation made by using our deduced consumer properties hascorrect rate26.28%.(3) Person recommendation made by using consumer past personal interest hascorrect rate12.21%.