Social Relationship Mining And Itinerary Recommendation System Based On Mobile Network Data

With the rapid development of Internet and mobile communication technology,a large quantity of digital footprints provides unprecedented opportunities for the complex structure and dynamic analysis of social systems.Mobile network data provides the basis for human sociality and mobility analysis at a macro scale,which can reflect the rhythm pattern and interest preferences of users' lives accurately.In recent years,mobile network data has been gradually applied to the tourism service industry due to the "group wisdom" contained.Mining travel routes and travel preferences of tourists from mobile network data can provide a reference for the construction of tourist attractions and surrounding service facilities.This paper designs and implements a Hainan smart tourism portrait system based on the Spark distributed platform.This system uses mobile network data and POI data to implement functions such as mobile network data cleaning,staying-areas mining,social relationship inference based on user mobility,and user behavior analysis.On this basis,the system constructes a knowledge graph of the tourism industry in Hainan.Combined with a large number of trajectories of tourists,it implements a platform with tourist portrait,attraction portrait and journey planning.The work of this paper is mainly reflected in the following aspects:First,we propose mobile network data cleaning algorithm and staying-area mining algorithm based on the Spark distributed platform.Then we statistically analyze the daily travel mobile behavior patterns of users in this province.Second,we conduct research on inferring social relationships based on the mobility of mobile network data.We define time parameter and location parameter to identify the meeting events of users,and then we build a mobility relationship network considersing the social relationship propagation characteristics.At last we use graph embedding algorithm to infer the hidden social relationships with POI data and user's residence and work place information.Third,we implement a next spot recommendation algorithm based on multi-peak capture.we first build a Hainan's tourism knowledge graph,and then builds a spots' transfers graph based on tourists' routes in mobile network data.The recall module of our recommendation system uses the LSHForest algorithm which provides two recall schemes based on user preferences and group wisdom.The refined ranking model introduces the user's social relationship and attention units to rank the recalled attractions collection.Finally,we build a smart travel portrait system of Hainan province.System integrates multi-domain data processing schemes and algorithms such as residence and work places identifying,inferring social relationships with mobility,constructing a Hainan tourism knowledge graph,and next attraction recommending.A web interaction platform based on bootstrap and tornado are also implentment for portraits of tourists and attractions showing and journey planning.