The Research And Implementation Of Search Strategy Evaluation Model

One of the purpose of search engine operation is to optimize the search engine,with the accurnulation of data and related technology.The operational work changes from the artificial decision-making to data-driven.As an important feedback mechanism in the operational work,the evaluation of search strategies also needs to be changed to meet new development requirementsIn the context of the explosive growth of Internet data,the Internet product's fast iteration and short development cycle and the increasement of low-quality cheating sites.How to provide a search strategy evaluation evaluation model with low cost of manpower,high time efficiency,accurate assessment,the authority of the site assessment is particularly important.This thesis studies search engine relevance computation model and authority computation model related characteristics,combined with A/B online experimental fiamework.Proposed for searching strategy evaluation model,designs and implements a prototype system.Thesis's main works include:(1)In order to solve the problems that dificulty of obtain labeled data,the time cost is large and the evaluation conclusion is not accurate.of the existing relativity evaluation methods.Based on the improvement of click model,modeling user click and browsing behaviors by presenting the assumptions.RelevanceMetric--the relativity evaluation algorithm suitable for A/B experiment is designed.(2)For the problems of content-based authority calculation model's time complexity is high and the authority evaluation of the site depends on manual judgment,AuthorityMetric--the authority evaluation algorithm for A/B experiment is designed.Improve the TrustRank algorithm,which highlights the authority of the theme-related features and shorten the scope of operation.According to methods that Combined with the query segmentation and term importance computation,in the premise of ensure the user's main requirement of query,effectively shotern the text matching time.(3)The experimental results show that the algorithm has good performance of accuracy,stability and efficiency in terms of precision,recall rate,F1 value and MRR.It can be used to meet the demand.(4)Design and implemente of prototype system based on the idea of hierarchical architecture and Hadoop,MySql technology.