User Interest Clustering Using Basic Attribute Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current cluster methods based on user interest fail to accurately determine user interest when no behavior data is available, leading to inaccurate content recommendations for new users.
Innovation Solution
A method and apparatus that calculate a distance between basic attribute data and a cluster center point using pre-determined data, comparing it with a predefined threshold to determine user cluster membership, allowing for content recommendations based on registered user information before behavior data is collected.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If cluster operation is performed based on user behavior data, then recommendation accuracy is improved for users with behavior data, but new users without behavior data cannot be clustered and receive inaccurate recommendations
Solution Approach 1:
The patent performs cluster operation in advance using basic attribute data obtained during user registration, before any behavior data is generated. This preliminary clustering enables new users to receive accurate recommendations immediately upon registration, eliminating the cold start problem while maintaining high measurement precision for existing users.
2Device complexity
If only web page behavior data is considered for user interest, then cluster operation is simple, but user interest cannot be determined for users without behavior data
Solution Approach 1:
The patent makes the clustering system universal by enabling it to process both basic attribute data (for new users without behavior data) and behavior data (for existing users). The system automatically selects the appropriate data source based on user status, ensuring reliable user interest determination across all user types while maintaining operational simplicity.
Data Source
AI summary
A cluster method and apparatus based on user interest are provided in the present invention. The method includes: receiving pre-determined basic attribute data of a user; calculating a distance between the basic attribute data and a cluster center point, based on the cluster point of the basic attribute data related to the calculation of user interest; comparing the distance with a predefined threshold; if the distance is less than the predefined threshold, determining that the user belongs to a cluster of the interest. As the embodiments of the invention take the basic attribute data from the data of registered users into consideration, the recommendation of corresponding content is executed before getting user behavior data of browsing operations, based on the cluster of interest obtained from register information. Thus the accuracy of recommending to new registered users may be improved.