With the rapid development of social networks, microblog has become a platform for the spread of information dissemination. In view of the characteristics of microblog information diffusion, mining the potential law of microblog forwarding process is of great significance for public opinion monitoring, hot topic tracking, product marketing and so on. In this paper, the Sina microblog is used as the data source, the user and microblog content are used as the characteristic source. The neural network prediction algorithm in machine learning is introduced to establish the prediction model to realize the forward prediction of microblog. The results show that, compard with the traditional prediction method, the microblog forwarding can make a higher accuracy prediction.

In view of the limitation of DL OWL reasoning and expression ability, a reasoning method based on first order logic FOL is proposed, and the method is applied in the personalized E-Learning areas. The E-Learning domain ontology library was built to translate the OWL DL ontology into the FOL ontology, and the FOL to implement personalized matching rules was used for E-Leaning resources. Finally, the first order theorem is used for reasoning, and the result of the reasoning of the three first-order theorem is compared. Experimental results show that complex reasoning of OWL DL ontology through the FOL is feasible. The proposal achieves a good trade-off between expressive and reasoning.

Based on the traditional electronic commerce recommendation system, these technology means such as distributed computing, publish/subscribe based on the semantic，and ontology were introduced. The personalized electronic commerce recommendation model architecture was designed. The film areas ontology was constructed, which was implemented by Protégé 3.3.1 ontology model tool. The architecture core algorithms were described in detail, which are the clustering algorithm of users characteristics and the match algorithm of publish/subscribe based on the semantic. The experimental results show that the personalized recommendation system based on publish/subscribe has a good recommendation accuracy.