Font Size: a A A

Prediction Of Moving Objects' K-Nearest Neighbor Based On Fuzzy-Rough Sets

Posted on:2009-03-11Degree:MasterType:Thesis
Country:ChinaCandidate:Y YuanFull Text:PDF
GTID:2178360245496470Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
The research of direction and neighbors' prediction of moving objects are challenged by there application on traffic attemper, service based on position, weather forecast, aptitiude navigation system, geography information system and military affairs. We could note and deposit moving objects' status messages with the help of GPS on the moving objects. Prediction of the future can be done by these postural data. Meanwhile, as a result of the influence from self-inflicted or outside factors, moving objects' status messages would change constantly. So there would be many uncertainties appear in the prediction. Many scholars have taken up with the disposals of the uncertainties that exist in the moving objects and its k-nearest neighbors' prediction. The uncertainty includes temporal one and spatial one. In this paper we mainly discuss the space-uncertainties of the moving objects' future direction.The technologys of neighbors' prediction are mainly composed by the technology of space data index and the arithmetic of neighbors' prediction based on index. Compared with others, TPR tree has more superiority on space of index, plan of data partition and index having seasonal rebuild or not. TP query could find neighbors of current point, and involves period of validity of the ends and the changes after this time. The uncertainty analysis of moving objects' neighbor prediction would be based on the index structure of TPR tree and the arithmetic of TP query. And then, compare and analysis by fuzzy rough sets will be used on the results of them.Have an analysis on the fuzzy membership degree of moving objects' predict position, which results from TPR and TP, as the uncertainty of its predicted position appears as the fuzzy membership degree that the actual position at the region around predicted position.Find its extended k + m nearest neighbor set, which is also educed from traditional methods.On account of the fuzzy-uncertainty of moving objects' predicted position, the accuracy of the k-nearest neighbor that obtained by the tradition KNN method would be affected. The rough-uncertainty of the k-nearest neighbor based on the indiscernibility relation of "k-nearest neighbor" emerges accordingly. the fuzzy-rough membership function is employed to obtain the final k-nearest neighbor set. At last, some dynamic data are produced by Genarate spatio temporal data. With TPR tree index for moving objects and TP query with infection time, we use the method of prediction of moving objects' neighbor by fuzzy rough sets to analyze the precision of the results. It is concluded that, compared to the actual position of the moving objects, the analysis based on the theory of fuzzy-rough sets can promote the precision of its k-nearest neighbor set noticeably, though we have enhanced the calculational quantity. The distance from each point in k + m nearest neighbor W to 2n sampled points that moving object P may lies in fuzzy space R, which is referred in fuzzy-rough membership function, has used sample method. It means that there may be an impact on the accuracy. So in future work, we would investigate more precise algorithm for the prediction of moving objects' k-nearest neighbor based on non-sampled method.
Keywords/Search Tags:Moving object, K Nearest neighbor query, Fuzzy sets, Rough sets
PDF Full Text Request
Related items