Analysis And Mining System Based On Massive Data

In the age of the electronic commerce,the electronic commerce, logisticsindustry, logistics vehicles are more prosperity than ever.More and more logisticsvehicles GPS data are produced.These data contain a lot of traffic information suchas road conditions,vehicles,and even social and economic development.Throughstatistics and analysis of vehicle driving distance, time, location, vehicle parkingcharacteristics, trajectory data mining can find shipping line characteristics,providelogistics company based on vehicle scheduling schemes such as time, cost, andderived a series of LBS application.Taking massive GPS data as the data source, using massive trajectory datamining and related theory of road recommending, composed by online and offlinesystem, this paper proposes and realizes a designframework of route recommendingsystem for logistics vehicle through establishing clustering model and analyzingmassive GPS data to understand the driving rule of logistics vehicles. The keyapproach is deep studying on data preprocessing, stops detecting, route segmenting,similar freight trajectory clustering and freight lines recommending. The specificwork is as follows:As a necessary work in trajectory data mining，i study the pretreatment method,including data cleaning, data of abnormal detection and exclusion, and with thecharacteristics of this system all the GPS data in analysis and put forward a kind ofanomaly detection algorithm based on the historical trajectory data. The algorithm formassive data processing has low time complexity.Parking points detection and path integral can find that the pattern of logisticsvehicles and goods.In this paper, on the basis of naive bayes algorithm, i put forwarda new way for trajectory segmentation,according to the logistics vehicle parking andordinary ponits using the different attributes of time and space between them whengoods are loaded and unloaded.I regulate freight trajectories clustering similar to the same starting point and endpoint of the trajectory. then project them on the same latitude.After that usingk-means algorithm, the characteristic of trajectory is analyzed, and finally get logistics vehicle general movement tracks.In the recommendation of shipping lines, based on the difference of historicaltrajectory data in time, distance and cost,i design and draw the correspondingrecommended route guidance which logistics driver adopts through reasonabledriving scheme.Compared with traditional pretreatment method.These tests show that themethod of pretreatment trajectory is faster, more efficient, but sacrificing someprecision. Detecting parking spots and track segmentation achieve good effects. In thecases of missing stops of vehicle trajectory analysis,the research results have veryimportant theoretical significance.