Improving the Support System of Public Sports Facilities Applying Text Mining and Multiple Focused on the support facilities for the National Sports Promotion Fund

Abstract

This study used text mining to analyze the user complaints about public sports facilities supported by the Korea Sports Promotion Fund and seek measures for improvement. Methods/Statistical analysis: A framework for sports texts should be designed to include diverse features for collecting and analyzing sports-related texts. Among other methods of topic modelling, this study used the most widely used probability model, LDA(Latent Dirichlet Allocation). Word2vec models are applicable for different purposes. This study used Word2vec to identify key words highly associated to relevant key words. Findings: The analysis highlighted the following. First, the LDA topic clustering analysis by type identified 4 important key words (instructors, members, swimming and failure), which were in turn explored further with Word2Vec. Second, the analysis of associated words found such salient words as swimming, members, time, center, class and fitness acceptance in relation to the general type, whereas members, swimming, time, center, exercise, class and lesson proved important in the complex type. Third, as for the frequency of words, swimming, members and center frequently appeared in the general type in the order named, whereas the complex and gymnasium types were associated with the importance of swimming, members and time, in the order named. Improvements/Applications: The present findings may serve as a guideline for public sports facilities as public goods to improve the quality of service for users based on the user complaints.