Generating Hypotheses Based on the Table Constraint Satisfaction Methods in JSM-Systems

It is known that constraint satisfaction methods are successfully used to solve many complex combinatorial search problems. As their distinctive feature, it should be noted the widespread use of logical inference procedures on constraints that implement the reduction of the search space. The article continues a series of works that deal with the development of table constraint satisfaction methods to solve data mining problems. Previously, the author’s methods of inference on table constraints for clustering problems, discovering patterns of the required type, and searching for association rules were presented. In this work, using the example of solving the binary classification problem, the possibilities of applying the author’s approach to modeling JSM-reasoning are considered for the first time. The article considers the case when the properties of objects are atomic and have no internal structure. Within the framework of the approach, positive and negative examples are presented using specialized table constraints, namely compressed tables of the D-type, and the process of generating JSM-hypotheses is performed using the rules developed for reducing the search space. The proposed method of generating JSM-hypotheses makes it possible to effectively solve high-dimensional problems.