Frequent Pattern Mining (FPM) is the most focused research by numerous researchers to predict the repeated activities happening on different sectors. The FPM is done utilizing Lasso Regression based Improved Frequent Pattern Mining Detection Scheme (LR-IFPMDS) in our previous research. However, in the existing work, performance degradation occurs at the time of attribute selection with multiple objectives. It reduced the overall prediction accuracy. Therefore, this work focussed on to overcome the existing work limitation and introduced Linear Prefix Tree based Frequent Pattern Mining Detection Scheme (LPT-FPMDS) for mining frequent patterns. In this work initially, optimal attribute selection performed using Hybrid crow swarm Optimization with Cat Swarm Algorithm. Here fitness values considered are Accuracy, Error rate, Information Gain and Gain Ratio. Based on selected attributes frequent pattern rule generation is performed. Rule pruning process is executed utilizing Laplace accuracy. In this work, rule pruning and frequent pattern mining scheme is combined together to obtain the frequent patterns. The proposed FPM technique uses Linear Prefix Tree (LPT) for recognizing the frequent patterns. In LPT, once the leaf node reached, it estimates the Laplace accuracy and executes the rule pruning procedure based on the estimation outcome. This enhances the accuracy and curtails the computation overhead. The entire work is executed using Matlab simulation tool. The experimentation result shows that the proposed model has performance improvement in terms of accuracy.

Pattern rules, frequent pattern, Linear Prefix Tree, Laplace accuracy, leaf node of tree.