An incremental model and an iterative model based on MaxSAT for learning interpretable and balanced classification rules

Abstract

The increasing advancements in the field of machine learning have resulted in the development of numerous applications that effectively address a wide range of problems with accurate predictions. However, in some cases, accuracy alone may not be sufficient. Many real-world problems also demand explanations and interpretability behind the predictions. Therefore, one of the most popular interpretable models for solving this type of problem is classification rules. This work aims to propose an incremental model and an iterative model for learning interpretable and balanced rules based on MaxSAT, called IMLIB and I-IMLIB, respectively. This new model was based on two other approaches, one based on SAT and the other on MaxSAT. The former limits the size of each generated rule, making it possible to balance them. We suggest that such a set of rules seems more natural to be understood compared to a mixture of large and small rules. The second approach, called IMLI, presents two techniques to increase performance. One technique involves learning a set of rules by incrementally applying the model to partitions of the dataset. The other consists of learning a new rule at each iteration in the incremental process of the first technique. The conducted experiments showed that IMLIB and its iterative version can generate smaller and more balanced sets of rules than IMLI and its iterative version. Moreover, the new approaches manage to stay close to or even surpass IMLI and its iterative version in terms of accuracy. Regarding training time, the iterative version of IMLIB achieved improvements compared to IMLIB, but the iterative version of IMLI still stood out in all tested datasets.