Users Relationship Strength Measurements And Analysis Based On The Heterogeneous Sensor Data

Recent years,with the replacement of traditional mobile phones by smart phones,which were equipped with an intelligent operating system(such as IOS,Android and Windows Phone)has become a mobile computing platform of communication,Entertainment,business office and computing.Various sensors are embedded in the smartphones,so we can collect the data that can reflect the hidden social relations between the users such as user's location,communication records,SMS records,daily trajectories.It is great important to develop the social networks and explore the development and evolution of social group structures by find out the relationship strength among people.Most of researches on social relationships and relationship strength between two users are based on social-relations data(social-relations data means call records,SMS records,the interaction of social network software,etc.).In this research,we propose a RSCHD(Relationship Strength Calculation based on Heterogeneous Data),which focus on non-social data to analysis the relationship strength between people thought daily trajectories,WiFi data and Bluetooth data.Our works in this research and contributions are mainly reflected as follows:Firstly,we use timeslice kalman filtering algorithm to remove the noising points in GPS data;finding out the stay points and using the clustering algorithm to detect the clustering of GPS data or stay points by density and time,and marking the semantic labels with points;on this basis,we use Dynamic Time Warping algorithm to calculate the similarity between user space trajectories.As for semantic trajectories of users,we put forward a two dimensions method,one the one hand,we use quickly simhash algorithm to calculate user‘s strength with semantic trajectories;one the other hand,after mining out the user's trajectory movement pattern,we can get the strength thought user's trajectory movement pattern.Thus we merge above dimensions.Secondly,the association topology is used to represent the WiFi sensors data at every piece of time.A new method is proposed to calculate the relationship strength based on WiFi data,we regard the similarity between two WiFi topology graphs as the similarity between different WiFi context information at the same time,which is the same as strength calculation via Bluetooth data.Finally,on the base of above results,ensemble learning is used to merge the measurement results of three levels as the final relationship strength between users.We develop a context information collection system StarLog to collected a long term sensors data of students to calculate the relationship between students,and effectiveness RSMHD for experiments,the results indicate that the RSMHD can effectively detect the relationship strength between users.