Mobility Pattern Mining Using Semantic Trees
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing mobility pattern mining methods focus on location-based data and fail to identify patterns related to topics, such as 'house decoration,' and are slow to recognize new patterns, which limits the quality of location-based services.
Innovation Solution
A device and method that collect historical data to extract semantic trees for specific topics, match stay points with leaf nodes, and determine candidate mobility patterns by combining sub-trees with topic-related stay points, selecting frequent patterns from a database based on semantic similarity and frequency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing methods focus on mining mobility patterns about locations based on frequency of visited locations, then location-based mobility patterns can be obtained, but mobility patterns about topics (e.g., house decoration) cannot be identified
Solution Approach 1:
The patent segments the mobility pattern mining process into two distinct components: location-based pattern mining and topic-based pattern mining. By separating these functions, the system can simultaneously handle both location frequency analysis and semantic topic recognition, allowing it to identify both traditional location patterns and new topic-related patterns without compromising either capability
Solution Approach 2:
The patent introduces an intermediary component that bridges location data and topic information. This intermediary layer processes location sequences and matches them with semantic trees to extract topic-related mobility patterns, enabling the system to translate raw location data into meaningful topic-based insights while preserving the original location-based analysis capabilities
2Reliability
If existing methods require related movements to achieve a certain number of times, then reliable mobility patterns can be identified, but new mobility patterns cannot be recognized early
Solution Approach 1:
The patent applies preliminary action by pre-building semantic trees and organizing topic-related location information in advance. When analyzing user mobility data, the system can immediately match observed location sequences against these pre-organized structures, enabling early detection of emerging mobility patterns without waiting for traditional frequency thresholds to be met
Solution Approach 2:
The patent changes the evaluation parameter from pure frequency-based metrics to a hybrid approach that incorporates semantic similarity and topic relevance. This parameter transformation allows the system to identify patterns with fewer occurrences by evaluating their semantic coherence and topic alignment, thereby reducing the time delay in recognizing new mobility patterns while maintaining reliability
Data Source
AI summary
A mobility pattern mining device includes: a data collecting unit configured to collect a user's historical data; a stay point obtaining unit configured to obtain stay points of the user from the historical data; an extracting unit configured to obtain a semantic tree for a certain topic and extract topic related stay points from the stay points of the user by using the semantic tree for the certain topic; a determining unit configured to determine a candidate mobility pattern of the user based on the topic related stay points; and a selecting unit configured to select a frequent mobility pattern which best matches the candidate mobility pattern of the user from a frequent mobility pattern database related to the certain topic, wherein the database is set in advance based on historical data for a plurality of users to obtain mobility patterns about topics for users.