Clustering for Complex Structured Data Based on Higher-Order Logic

Data clustering is an important technique for exploratory data analysis, and has been studied for many years. The existing clustering methods are all designed in attribute-value setting or first-order logic setting. However, attribute-value language can not describe complex structured data. First-order logic can represent certain complex structured data, but both scalability and efficiency of clustering algorithms in this setting are questionable because they need vast scans of data. This paper presents clustering for complex structured data based on higher-order logic. Data is represented by Escher, which is a typed, higher-order logic language. K-means algorithm is investigated with it. Experimental results demonstrate that clustering algorithm which adopts Escher have higher efficiency and better scalability.