A state-of-the-art prefix-based frequent pattern mining without candidate generation and compact FP tree generation

Without the candidate generation approach, it is still dominating and gaining a good research impact to find the desired association rules. The FP tree is a memory resident that sometimes memory overfits for high-volume datasets. The issue with the FP growth deals with numerous pointers. It generates a massive number of conditional pattern base and conditional FP trees that pursue notable performance degradation with specific datasets. FP growth needs to maintain many pointers operations for large datasets during the rule mining process. We present an efficient frequent patterns approach known as prefix-based frequent pattern mining (PBFPM). A straightforward novel array-based key-value pair approaches for finding frequent patterns efficiently from large-volume datasets. We induce an array structure table (AST) rather than an FP tree structure for storing the dataset's pattern. The proposed method does not generate duplicate frequent patterns and avoid numerous pointer dealings, which saves time in the rule-generation process. We compared the performance concerning time and memory complexity with the FP tree and state-of-the-art boss tree.