Knowledge Discovery in a Facility Condition Assessment Database Using Text Clustering

Knowledge discovery in databases (KDD) has been applied in many different areas of study including DNA sequence analysis, pattern discovery, document classification, image recognition, and speech recognition. This paper presents the application of KDD in the analysis of a facility condition assessment (FCA) database. The FCA database contains information on facilities located at three campuses within a statewide university system. The case study utilizes cluster analysis for text mining. Cluster analysis is the grouping of objects that are similar within the same cluster and dissimilar to the other clusters. In this analysis, deficiency descriptions from a university’s FCA database are the objects being grouped together into clusters. Deficiency descriptions were gathered from 15 housing facilities and 15 academic facilities located at 3 campuses. The results show how some clusters of facility deficiencies are unique with respect to the type of facility and the influence of location on deficiencies of academic facilities. The paper begins with a presentation of background on clustering approaches in KDD. Next, a case study based on a higher education FCA database is presented. Last, the paper concludes by exploring other potential areas of application of the described clustering approach.