Geo-referenced Trajectory Clustering via Fréchet Distance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data track mining algorithms for geo-referenced trajectories require time-consuming user interaction and often produce artifacts due to the use of metrics that do not satisfy the axioms of a metric space, particularly when determining the number of clusters and selecting preferred routes.
Innovation Solution
The implementation of a system that automatically determines the number of clusters in geo-referenced track data using a Fréchet distance metric that satisfies the triangle inequality, based on the Minimum Description Length (MDL) principle, allowing for clustering without user input and reducing artifacts by employing a farthest point clustering heuristic to minimize the maximum diameter of clusters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If conventional clustering algorithms are used, then clustering can be performed, but user interaction is required which increases time consumption and reduces productivity
Solution Approach 1:
The system automatically determines the number of clusters and performs clustering without requiring user interaction. The algorithm self-adjusts parameters and selects the optimal number of clusters based on the data characteristics, eliminating the need for manual user input while maintaining clustering effectiveness
Solution Approach 2:
The system dynamically adjusts clustering parameters including the number of clusters based on data analysis. By changing parameters automatically rather than requiring fixed user input, the system adapts to different datasets and achieves both ease of operation and high productivity
2Productivity
If conventional metrics are used for clustering, then clustering can be performed, but artifacts appear in the results which reduces measurement precision
Solution Approach 1:
The system changes the metric parameter from conventional distance metrics to Fréchet distance metric that satisfies metric space axioms. This parameter change eliminates artifacts in clustering results while maintaining computational efficiency, simultaneously achieving both productivity and measurement precision
3Extent of automation
If the number of clusters is not predetermined, then automation is improved, but determining the optimal number becomes more complex
Solution Approach 1:
The clustering algorithm automatically determines the optimal number of clusters by analyzing data characteristics and evaluating clustering quality metrics. The system self-regulates the cluster count without external intervention, achieving high automation while managing complexity through efficient algorithms
Solution Approach 2:
The system uses feedback mechanisms where clustering results are evaluated and used to adjust the number of clusters iteratively. By incorporating feedback loops that assess clustering quality, the system automates the determination process while keeping algorithm complexity manageable through structured evaluation criteria
Data Source
AI summary
Embodiments of a method and system for identifying clusters in collections of data entities are generally described herein. In some embodiments, the method includes defining a metric space over the data entities. A distance function of the metric space may satisfy the triangle inequality. The method may include determining, based on the distance function of the metric space, a value for a number of clusters that minimizes a number of data bits used to define a model of the collection of the data entities. The model may thereby describe the collection of data entities using a minimum description length (MDL). The method may include assigning data entities of the collection of data entities to the clusters. The number of clusters to which the data entities are assigned may correspond to the determined value.