Analysis Of Location Big Data Based On Unsupervised Learning Technique

With the development of society science and technology,location aware technologies such as mobile communications and sensing devices form a large number of location data,mining and analyzing the data of these positions,so as to find out the potential useful information,and make the urban construction planning and route planning reasonably,will greatly promote the intelligence and information of modern society.At the same time,it is also helpful for researchers to speculate on the ownership,income level and occupation of people's transportation vehicles from the trajectory data of people.In this paper,some related research on the location large data.First of all,the GPS trajectory data and the influence factors of the data error are analyzed,and to do the pretreatment on the original data,including error analysis of data,data processing basis and method,after that,the starting and ending points of the residents are identified,and the ArcGIS is used to visualize the map.Based on the processed GPS data,the temporal and spatial distribution characteristics of residents' travel are analyzed,including the number of working days and rest days,the peak time of daily travel,and the relevant conclusions are obtained.The cluster analysis method is applied to the research of taxi GPS data,and the reasonable clustering method is used to divide the traffic area of the study area,and the travel OD matrix is established.Secondly,using K-Means method based on spatial clustering,get the passengers to get off the active center,and combined with ArcGIS spatial analysis tools,building a buffer to find the easiest way for passengers to take a taxi,solving the problem of moving the center off the road.In addition,the density clustering method is used to study the distribution law of urban residents' hot spots.The effectiveness of the proposed method is proved by combining with the actual situation.Then,in order to improve the recognition rate of different travel modes,this paper proposes a method of deep learning to identify the travel patterns,based on the data collected from the GPS by Microsoft Asia Research Institute,the characteristics of different travel modes are analyzed by using the time characteristics of different travel modes,considering the influence of the number of iterations on the root mean square error and the training time of the network,we choose the appropriate number of iterations to achieve the best training effect in the shortestpossible time.Finally,the method is compared with BP neural network and SVM method,the results show that the model proposed in this paper can be used to identify the different travel modes effectively,the accuracy of the method is obviously improved compared with the traditional method,and the feasibility of the model is proved.