A Rule Measure to Represent the Temporal Changes of Data Mining Patterns

Data mining is the automated extraction of hitherto unknown patterns in large databases. Patterns or rules extracted by mining algorithms are generally true for the current state of the database. As the database state changes due to new transactions these rules may either become invalid or may have increased support. In this paper we present a new rule measure – Rule Trend Analysis (RTA) – which quantifies and represents the trend or behaviour of patterns extracted by data mining systems as these rules evolve along a continuous time interval. This measure is derived using the database log file by analysing transactions that are compliant with the rule and those that are not. We present both the theoretical aspects as well as the results obtained by implementing this technique of data mining rule analysis.