An enhanced intelligent database engine by neural network and data mining

Abstract

An Intelligent Database Engine (IDE) is developed to solve any classification problem by providing two integrated features: decision-making by a backpropagation (BP) neural network (NN) and decision support by Apriori, a data mining (DM) algorithm. Previous experimental results show the accuracy of NN (90%) and DM (60%) to be drastically distinct. Thus, efforts to improve DM accuracy is crucial to ensure a well-balanced hybrid architecture. The poor DM performance is caused by either too few rules or too many poor rules which are generated in the classifier. Thus, the first problem is curbed by generating multiple level rules, by incorporating multiple attribute support and level confidence to the initial Apriori. The second problem is tackled by implementing two strengthening procedures, confidence and Bayes verification to filter out the unpredictive rules. Experiments with more datasets are carried out to compare the performance of initial and improved Apriori. Great improvement is obtained for the latter