Land Use Classification And Analysis Based On Cell Phone Call Data

With the development of economy and society and the advancement of science and technology,the way of land use is constantly changing.The classification of land use is the basis and prerequisite for determining the current status of land use,which mainly reflects the forms and impacts of human beings on land use,there are different parttens of human activities in different land use classifications.The type of land use can be determined by the surface physical features of the surface features and the functional division of the community.Traditional social survey methods have high costs,small amounts of data,and long periods;remote sensing technology is considered an important means of land use classification,but remote sensing data can only capture the physical features of surface features and cannot be identified by residential activities within the city.Human activity pattern.The development of an information society has produced a great deal of data related to human activities.Due to the widespread application of mobile phones and the rapid development of communication networks,residents' activities can be retrieved from mobile phone data to indicate the social functions of land use.In this study,the mobile phone call data of Milan City,is used as the data source.Decompose time-series decomposition algorithm is used to decompose the cell phone call sequence data into trend items,seasonal items,and random items.Firstly,based on the decomposition algorithm,the spatiotemporal characteristics of mobile phone call data are analyzed,in the feature of time distribution,the cyclical change of mobile phone call data within 24 hours and the trend change characteristics within one week are analyzed.In the spatial distribution feature,spatial clustering based on Kmeans clustering algorithm and normalized value-based Nuclear density analysis.The study found that cell phone call data is representative of human activity intensity and urban spatial form.Based on the supervised random forest classification algorithm,three feature extraction methods are used:Residual,TrendSeasonal,and Random.Cell phone time series call data is characterized for land use classification.The study found that the use of random items as the feature point for the classification of land use is the highest accuracy,up to 55.22%.At the same time,this study is based on Twitter's AnomalyDetection time series anomaly detection algorithm to detect and correct mobile phone call data sequence time series outliers.It is found that the accuracy of the time series data after the outliers are corrected for land use classification is reduced.Finally,it is found that the accuracy of land-use classification based on mobile phone data is relatively low compared with domestic and foreign research.he reasons for the lower classification accuracy and the prospects for follow-up work are analyzed.This study provides a new research perspective and ideas for land use classification research based on mobile phone call data.