In practical scenarios, an entity presence can depend on existence probability instead of binary situations of present or absent. This is certainly relevant for information taken in an experimental setting or with instruments, devices, and faulty methods. High-utility patterns mining (HUPM) is a collection of approaches for detecting patterns in transaction records that take into account both object count and profitability. HUPM algorithms, on the other hand, can only handle accurate data, despite the fact that extensive data obtained in real-world applications via experimental observations or sensors are frequently uncertain. To uncover interesting patterns in an inherent uncertain collection, potential high-utility pattern mining (PHUPM) is developed. This paper proposes a Spark-based potential interesting pattern mining solution to work with large amounts of uncertain data. The suggested technique effectively discovers patterns using the probability-utility-list structure. One of our highest priorities is to improve execution time while increasing parallelization and distribution of all workloads. In-depth test findings on both real and simulated databases reveal that the proposed method performs well in a Spark framework with large data collections.
原文链接
分享
引用
