Optimal Search-based Distributed Data Retrieval Technology

With the rapid development of internet and the increase of information in internet, it seems impossible for users to search the information they need from large mount of data without powerful retrieval and analysis tools. At present, the widely used retrieval system can solve the problem of resource discovering partially, However, with the rapid expansion of network data resources, traditional C/S retrieval method has been unable to meet the user's requirements in performance of search. When the scale of information retrieval reaches a certain extent, it is necessary to adopt distributed approach to improve system performance.Bringing forward technology of distributed data retrieval is of great significance to the field of information retrieval. Compared with the traditional information retrieval technologies, the efficiency of search retrieval system which uses distributed data retrieval technology is improved significantly. Then how to improve the retrieval efficiency on the basis of distributed data retrieval technology is a topic worth researching. Optimal search theory developed during World War II, it is a branch of statistical decision theory in operations Research which studies how to distribute resources in order to search target with biggest probability and smallest consumption of resources. We used optimal search theory on distributed data retrieval to get the results of optimal the quality of retrieval system.The knowledge of distributed data retrieval technology and optimized search theory are investigated in this dissertation, and the combination technology of above two theories is researched, especially. At first, theory of optimal search is used to establish mathematical model of the distributed data retrieval system, analyzing initial probability distribution of optimal search model and determining the form of detection function, then the strategy of optimal allocation could be instituted in the condition of limited time, which maximizes search probability of the target and minimizes expectations of the retrieval time in the condition of user's search for fixed results. Finally, the essays studies how many errors the initial probability distribution would cause to detection probability. In addition, in accordance with the model of Distributed Data Retrieval System ,the dissertation designs and implements the system on the base of theory of optimal search ,on the base of which a series of experiments have been carried out , discussing how the retrieval order of system and initial probability distribution pact on system performance. Two evaluations of the system have been made through the aspects of precision ration and searching time, confirming the theory of optimal search has obvious optimization effect on Distributed Data Retrieval System.