A Pivotal Technology Research Of OLAP Session Recommendation Based On Multi-objective Optimization

OLAP(Online Analytical Processing),which is a core technology in business intelligence field,has occupied an important role in massive data analysis and aided decision support.While recommending for OLAP queries could help analyzer acquire the desired results quickly and explore the latent value of data in order to raise the decision efficiency.In view of the problems of the obscure subject and the low effect recommend method,the OLAP session is first regarded as the recommended object to enhance the topicality in this paper,then we study the core technology of OLAP session recommendation from the three objectives which contains similarity calculation,recommend method and tag technique,after which we optimize the three ones and finally a series of experiments are conducted to confirm the efficiency and rationality of our proposal.In detail,the innovation point and main work of this paper can be summarized as follows.Firstly,research on the features of OLAP query and OLAP session.The OLAP query is replaced with the OLAP session to be the basic unit of recommendation in order to raise the purposiveness and topicality of analysis.In addition,we introduce the improved Smith-Waterman algorithm to optimize the similarity detect approach,which can well raise the precision and recall of detection.Secondly,aiming to the cold start problem in collaborative filtering recommendation,the recommendation method which combines content-based recommend and collaborative filtering recommend is proposed in this paper,which adopts the content-based method in head-session and collaborative filtering method in end-session.In the meanwhile,we add the novelty ranking to make the result of recommend to be more timeliness,and finally with the help of Top-K recommendation,the diversity of recommend has been enhanced to find out the valuable information for users.Thirdly,the edit-based tag generation method is put forward to add the tag to the recommended session in order to cope with the specialty issue in OLAP field.Meanwhile,we use the edit-based method to match the recommended session with the similarity session that has a tag and sign the difference while in recommendation tosolve the sparsity of tags,which could enhance the comprehension of the newcomer.Fourthly,in order to confirm that our similarity calculation and recommend method has higher precision and recall together with a faster execution speed in our tag generation method,we have conduct a series of experiments using open source software and datasets,the result of which shows that our similarity calculation has about 15% higher than the traditional cosine vector method,the recommend method this paper has put forward ranges from 5% to 20% higher than the collaborative filtering on precision,recall,coverage and the other two measures,in addition,our tag generation technology is nearly twice faster than tag cloud method,and is hundreds of times faster than cinecube.