Geographic Clustering via Grid Segmentation and Centroid Reconfiguration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional clustering models fail to accurately create non-overlapping clusters in dynamic geographic areas, leading to improper allocation and inefficiencies in real-time data distribution, particularly in services like food delivery where precise location changes are involved.
Innovation Solution
A method and system that create clusters by identifying reference points, forming a rectangular boundary, dividing it into blocks, reconfiguring blocks based on centroid calculation, and using the 80-20 hypothesis to mark blocks as fixed or floating, allowing for iterative adjustments to merge or re-mark blocks to ensure non-overlapping clusters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional clustering models are used to group customers into clusters, then customer segmentation can be performed, but overlapping clusters are created leading to improper allocation and delivery failures
Solution Approach 1:
The patent divides the service area into discrete, non-overlapping grid cells as basic units. Each grid cell is independently classified as belonging to a specific customer cluster or remaining unassigned, eliminating the overlap problem inherent in conventional continuous clustering methods. This segmentation approach ensures clear boundaries and unambiguous allocation of delivery zones.
Solution Approach 2:
The patent applies different properties to different parts of the service area by classifying each grid cell individually based on its specific characteristics and distance to customer locations. This local classification approach allows the system to optimize cluster assignments for each specific area while maintaining non-overlapping boundaries, rather than applying a uniform clustering method across the entire region.
2Productivity
If conventional clustering models are used for real-time data distribution, then customer segmentation is achieved, but location imprecision of moving vehicles leads to ambiguous clustering results
Solution Approach 1:
The patent segments the continuous geographic space into discrete grid cells, which provides a natural tolerance for location imprecision. Since each grid cell represents a defined area rather than a precise point, small errors in vehicle location measurement do not lead to ambiguous cluster assignments. The discrete nature of grid cells makes the system more robust to measurement errors compared to continuous clustering methods.
Solution Approach 2:
The patent预先 creates a buffer against location imprecision by using grid cells of sufficient size that can accommodate minor location errors. The grid cell approach inherently cushions the system against measurement errors, ensuring that even if vehicle location data is slightly imprecise, the cluster assignment remains accurate and unambiguous.
3Ease of manufacture
If conventional clustering models create clusters without accounting for dynamic location changes, then initial clustering can be performed, but real-time clustering becomes challenging and inaccurate
Solution Approach 1:
The patent enables dynamic cluster assignments by allowing grid cells to be reclassified in real-time as vehicles move and customer locations change. The system can efficiently update cluster assignments by evaluating which grid cells should be reassigned based on current conditions, making the clustering adaptive to dynamic environments while maintaining the computational simplicity of the grid-based approach.
Solution Approach 2:
The discrete grid cell segmentation facilitates real-time updates by allowing individual cells or small groups of cells to be reassigned without recalculating the entire clustering structure. This modular approach maintains implementation simplicity while enabling accurate real-time clustering responses to changing conditions.
Data Source
AI summary
A system and a method to create clusters in an area are disclosed. Initially, a set of location data points associated with a plurality of user devices in an area and a defined block area are received. Further, a rectangular boundary is created by connecting a first reference point, a second reference point, a third reference point, and a fourth reference point identified based on a latitude and a longitude of each user device. Furthermore, a set of blocks are created by dividing the rectangular boundary based on the defined block area. The set of blocks are sorted based on a number of location data points present in each block. Subsequently, the set of blocks are reconfigured by determining a centroid of a plurality of location data points present in each block. Finally, a plurality of clusters is created in the area upon marking the reconfigured set of blocks.