Fuzzy Association Rule Mining for the Analysis of Historical Process Data

Abstract

Process data collected during the operation of complex production processes can be used for system identification, process monitoring and optimization. This work presents a new algorithm that is able to extract useful knowledge from data. The extracted information is given in the form of association rules. Association rule mining finds interesting association or correlation relationships among a large set of data items. The large itemsets can be related to the frequent events of a process, and this is useful for detect unknown relationships among the process variables, reduct the models of the system, estimate the product quality and build a classifier. The proposed method based on the Apriori algorithm, but the main idea is incorporate fuzziness (fuzzy logic increases the interpretability of the model and tolerance against measurement noise and uncertainty). The general applicability and efficiently of the developed tool are showed by an application study, one general example for the feature (input) selection problem and the analysis of a polymerization process data. Moreover the proposed classifier is used for three general used classification problems.