Answers are grounded in this paper and cite their sources.
Objective: In order to apply association rule mining to data sets with continuous variables, it is necessary to convert the variables into categorical structure. Therefore, we aim to compare the results obtained by categorizing continuous variables using different methods and analyzing them with association rule mining. Material and Methods: In this study, ChiMerge, clustering, Minimum Description Length Principle, equal interval, equal frequency methods were used to transform continuous variables into categorical structure by discretizing them. Various datasets were generated in the R with sample sizes of 100, 200, 500 and one binary dependent variable was created in each dataset. Additionally, 2, 3, 4, 5, 6, 7 continuous variables with a standard normal distribution were generated and various methods for transforming variables into categorical format were applied. A support of 10%, a confidence level of 80% were used. The number of rules varies based on the number of variables and the number of categories. Results: The study's results compared the descriptive statistics of the number of rules and the lift values. It can be said that high lift values are observed in scenarios with high levels of correlation and a higher number of variables and increasing the sample size can reduce the lift values of rules. Conclusion: In terms of the number of rules, the ChiMerge is the most affected by increasing the sample size. Furthermore, the ChiMerge yields stricter and higher lift values compared to other methods. While using association analysis, data type and multi-level associations should be considered.
No takes yet. Share an insight, caveat, or question.
SUCU et al. (2024) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: