A novel trend prediction system design via association rule mining represented in knowledge graphs: BIM research field application

Objective: A large number of scientific research and the rapid developments in the field of science and engineering make it almost impossible to analyze the literature manually. This study proposes a rule-based graph approach to analyze the knowledge areas and trend prediction for future studies. Material and Methods: In this study, a rule-based graph was created by utilizing the Apriori algorithm. In addition, keywords that have the same meaning but are important for the literature with a spelling difference were combined to obtain their exact values. Thus, the outputs obtained as a result are more understandable and accurate. BIM-related data were exported from the Scopus database as an exemplary research area. The Scopus database was used for this research because it is so comprehensive. The analysis was visualized with arulezViz and a more understandable result for users was aimed. Lift values of future trends were predicted with Random Forest, a machine learning technique. Results: The association rules between multiple word-groups of keywords allowed detailed and quantitative analysis inferences. Some words that mean the same thing but differ in spelling were manually combined into a single keyword. Some selected rules were entered into regression analysis by machine learning technique, and near future trends were determined by lift values. Conclusion: With this scientometric analysis approach, it is made possible to analyze the effectiveness of the keyword in the field of study in more detail. Not only the trends in the current BIM data, but also the trends in the near future have been shed light on. In future studies, the analysis can be enriched by adding different variables. Keywords: Building Information Modelling, Association Rule Mining, Knowledge Graph, Trend Prediction, Knowledge Management, Apriori Algorithm