Building associative rules for a database with a target parameter

A new efficient algorithm GoalApriori is proposed, which allows you to build associative rules for a special but important case when the original relational database has a target parameter. A classic example of such databases is medical databases, where the diagnosis made by doctors acts as a target parameter. Without loss of generality, we can assume that the target parameter is a discrete type parameter with a fixed set of values. The algorithm builds associative rules, the conclusion of which is the specific value of the target parameter. Premise of rules represents a set of properties of the database input parameters. The source database is reduced to a special format in which the resulting database record is specified as a single integer, regardless of the size of the source database record. In addition to saving memory, this format allows you to fully preserve information about the parameters representing the original record. More importantly, the computationally complex operations on records required to calculate the characteristics of rules are performed in this format almost instantly by a pair of logical operations on integers. The tasks and properties of the algorithm are considered. A number of statements regarding the properties of the algorithm are proved. The concept of a generalized criterion for the quality of rules is introduced, which allows for the ranking of rules.