Research And Application Of Text Recommendation Method For Experimental Detection And Network Communication Platform

Research materials in the field of petroleum exploration are of great scientific value,e.g.,field reports and research papers.Sharing it via the network platform for scientific researchers can help to improve the utilization of resources.With the increase of data year by year,this thesis focuses on information acquisition and carries out research on personalized recommendation system to screen and filter the data.In this thesis,an improved collaborative filtering method and a positive sample filling method are proposed to address the problems of sparseness,cold-start,and sample imbalance.And then,the proposed methods are employed to the experimental detection and network communication platform to design and implement the research materials recommendation subsystem.Firstly,for the sparseness,cold-start of research materials and sample imbalance problem of the Factorized Metric Learning recommendation method,a recommendation method based on textual features and distance factorization is presented,called RTDF.It utilizes textual topic features to construct the prior position of research materials to increase constraints on the model,and add a confidence term that considers user activity to the loss function.The experimental results show that the RTDF method can enhance the recommendation effect of research materials effectively.Secondly,to solve the low recommendation accuracy of low-activity users,a positive sample filling method based on tag information is proposed,called PFT.For low-activity users,the user's interest in research materials is calculated by the ternary relationship between users,research materials and tags,and employed to screen potential positive samples from unlabeled samples for filling.From experimental results,it can be seen that PFT can enhance the recommendation accuracy of low-activity users effectively.Finally,the proposed approaches are applied to the experimental detection and network communication platform to design and implement the research materials recommendation subsystem,which can help researchers to obtain potentially valuable information.