Knowledge-Based Interactive Postmining of User-Preferred Co-Location Patterns Using Ontologies.

Area of Science:

• Spatial data mining
• Geographic information science
• Database systems

Background:

• Co-location pattern mining is vital in spatial data mining.
• Large spatial datasets generate excessive co-location patterns, limiting practical use.
• Existing statistical methods fail to guarantee user-preferred pattern discovery.

Purpose of the Study:

• To develop an interactive approach for discovering user-preferred co-location patterns.
• To address the challenge of pattern redundancy in spatial datasets.
• To enhance decision-making through relevant pattern identification.

Main Methods:

• A novel interactive framework for user-guided co-location pattern discovery.
• Utilizing ontologies to quantify similarity between co-location patterns.
• Implementing a pattern filtering model to express user preferences and prune results.

Main Results:

• The proposed approach significantly reduces the number of co-location patterns.
• Filtered results are reduced to a few dozen patterns or less.
• On average, 80% of the selected patterns are preferred by the user.

Conclusions:

• The interactive approach effectively assists users in discovering preferred co-location patterns.
• Ontology-based similarity measurement and preference filtering enhance pattern relevance.
• This method offers an efficient solution for managing and utilizing large spatial datasets.