System and method for data mining from relational data by sieving through iterated relational reinforcement

摘要

A system and method are provided for performing the process known as "data mining" on a database of raw data records having common data elements, to obtain categorical cluster rules as to what elements of the data tend to occur in common in multiple records. Initial values are assigned to the elements. In an iterative process, the associated value for each given one of the elements is recalculated based on the values of other elements which occur in records together with the given element. Thus, the associated values will tend to grow for elements occurring together in multiple records. Those common occurrences of elements in multiple records represent categorical cluster rules the owner of the data is likely to want to know about. Thus, these rules may be identified based on the growth of the associated values for the records.