The Rules of the Intelligent Software Analysis Association

الملخص

This paper presents the concept of association rules through intellectual analysis with the aim of bug fixing. Finding the association rules allows determining the relations or connection between the specified values of categorical variables in large databases. This problem is often met in many projects on Data mining (knowledge discovery in databases process). Such research methods have many applications in many fields of business and research – starting from the consumer demand analysis or human capital management and up to the history of language. The association rules definition method is based on three statistic indicators, being calculated for pairs of objects (events, which happen simultaneously) in the data, named "Cause" and "Sequence"- Support (how often "Cause" and "Sequence" objects in the data are met, Trust (probability of the fact how often both objects "Cause" and "Sequence" among the data are met together) and Correlation (support for "Cause" and "Sequence", divided into the square root of the support result for "Case" and support for ‘Sequence"). The aim of the method is finding the type of the links "If "Cause", then "Sequence". The main problem of conformity detection methods is the enumeration of possibilities within an acceptable time. The known methods either artificially restrict such enumeration or build the decision trees, having principal search efficiency restrictions of the "If-then" rule. Other problems are connected with the fact, that known association rules searching techniques do not support the generalization function of the found rules and the optimal composition search function for such rules. A successful solution to the indicated problems can make a subject of new competitive developments.