User profile uses the techniques of data analysis and data mining to accurately describe the user’s characteristic attributes through the analysis and mining of their behavior data, which is widely used in personalized service and recommendation systems. User’s daily activity trajectory data contains information such as frequent places, places of residence, and work places, and is an important dimension for constructing user profile. In this paper, the clustering analysis method is used to mine the frequent locations of the user’s daily activity trajectory data, and a density-based spatio-temporal clustering algorithm T-DJ-Cluster is proposed to process the time and space dimensions of the data to make up for the shortcomings of traditional algorithms in processing the time dimension.
In this paper, the user profile constructed by the trajectory data mining method are used in the house exchange recommendation, and user is recommended to the listing information that meets their individual needs, thereby reducing the daily commuting time of them. In exchange process, this paper will design a graph-based similar user matching algorithm to achieve the matching process between multiple users. By looking up the user and location points and the connections between them into a directed graph, and designing a loop search algorithm in a directed graph to find candidate exchangeable users.
-
Development environment: macOS
-
IDE: PyCharm
-
Language: Python 3.6.4, JavaScript, Django 2.2.6
-
Databases: mysql 8.0.13, Neo4j 3.5.6
-
UserProfile: Main processing
-
cluster_method: T-dj-cluster Algorithm
-
preprocessing: Data preprocessing
-
data_analysis: Analysis of clustering results
-
0a0aaa2c5802f17d89ac93dc57633b76.csv : Test dataset
-
neo_FindLoop.py : Searching loops in directed graph in Neo4j