The Predictive Analysis Of Vehicle Demand Based On Multi-dimension Weighted Graph

With the continuous development of semantic web,originally simply connected data have formed a huge amounts of connecting data network with multi-dimensional weights.How to effectively deal with and integrate the data has now become the focus of current research.Aiming at the currently large amounts of related data and the complex attribute relationship among them,this paper proposes the concept of multi-dimensional weighted graph,which is to convert the RDF triple SPO data into a custom graph model V(point)E(line)W(weight),and thus convert a single RDF graph data into a multidimensional weighted graph.After the transformation,one of the variables is predicted by a weighted prediction function based on a time recurrent neural network and by connecting other associated variables in a multi-dimensional weighted graph.In addition,for the situation that the data of the weighted graph in real world is imperfect and the weights missing,this paper also proposes an information complement method based on KNN algorithm to improve the data.At the same time,the parallel processing of the shortest path is introduced in this paper,which effectively improves the data computing capability.Finally,in this multi-dimensional weighted graph model,with Large Knowledge Collider platform combined,this paper carries out related theoretical practice and draws a relative optimization strategy.To sum up,in the context of massively correlated multi-dimensional data,this paper proposes a data processing model based on multi-dimensional weighted graph,a weight prediction method based on a time recurrent neural network and an information complement method based on KNN algorithm for the bottlenecks that semantic web inference systems currently may encounter.Combining these three methods above with a flexible,open and scalable Large Knowledge Collider platform will surely provide heuristic help for the ever-changing semantic web data processing.