Social Relationship Detection from Network Activity Logs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods fail to accurately detect social relationships between users from diverse network services, often resulting in false positives or negatives, especially in high-density populations or geographically distributed groups.
Innovation Solution
A social relationship detection technique that analyzes user activity logs from network services over a prescribed period, utilizing timestamps, public network identifiers, and clustering factors to generate a data structure identifying social relationships, including family, friends, and colleagues, with the ability to distinguish deeper relationships.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing methods are used to detect social relationships from network services, then detection can be performed with basic data, but accuracy is low resulting in false positives or negatives
Solution Approach 1:
The patent changes multiple parameters simultaneously: temporal parameters (time of day, day of week, duration of co-occurrence), spatial parameters (geographic location, network identifier), and interaction parameters (type of network service interaction). By analyzing multiple parameters together rather than single parameters, the system achieves higher detection accuracy and reduces false positives/negatives in social relationship detection.
2Measurement precision
If basic user interaction data is used for detection, then data collection is simple, but detection confidence is low
Solution Approach 1:
The patent segments the social relationship detection process into distinct analytical components: temporal pattern analysis (time of day, day of week), spatial analysis (geographic location, network identifier), interaction frequency analysis, and co-occurrence duration analysis. Each segment processes specific aspects of user data independently, then results are combined to produce high-confidence detection outcomes. This segmentation manages complexity while improving detection confidence.
3Adaptability or versatility
If diverse network services are analyzed, then coverage is improved, but data heterogeneity increases making detection harder
Solution Approach 1:
The patent implements a universal detection framework that processes data from diverse network services (social media, messaging, gaming, retail) using the same analytical methodology. The system extracts universal features (temporal patterns, spatial information, interaction frequency) that apply across all service types, enabling consistent detection across heterogeneous data sources without requiring service-specific analysis logic.
Data Source
AI summary
Social relationships are detected between a plurality of users. A set of user activity logs is input, where these logs originate from one or more network services and cover a prescribed period of time. Each of the logs originates from a particular one of the services and includes a structured data stream that records each interaction that each of one or more of the users has with the particular one of the services. This structured data stream includes a timestamp specifying when this interaction occurred, and a public network identifier specifying a network address from which this interaction occurred. The set of user activity logs is then analyzed to detect a prescribed kind of social relationship between the users. A data structure is then generated that identifies the users between whom the prescribed kind of social relationship is detected, and the data structure is output.