Product Forecasting Based on Average Mutual Information and Knowledge Graph

The paper presents a method of modeling the training data which provided by China Conference on Knowledge Graph and Semantic Computing (CCKS) based on average mutual information and knowledge graph. Firstly, calculating the contribution of product attribute to the categories of product, and establishing the product prediction model of product. Then constructing the knowledge graph of training samples which is the network among attributes and categories of product; The average mutual information between attributes and categories is used to provide contribution value for the product prediction model, and the product knowledge graph limits the number of product categories effectively. This is an attempt to integrate algorithm of product forecasting with knowledge graph. After evaluating on the data released by CCKS2016, results show that classification model between average mutual and knowledge graph has high efficiency and accuracy.