Performance Analysis of Heuristic Miner and Genetics Algorithm in Process Cube: a Case Study

Abstract

Databases that are processed in the form of Online Analytical Processing (OLAP) can solve large query loads that cannot be resolved by transactional databases. OLAP systems are based on a multidimensional model commonly called a cube. In this study, OLAP techniques are applied in process mining, a method for bridging analysis based on business process models with database analysis. Like data mining, process mining produces process models by implementing the algorithms. This study implements the heuristic miner algorithm compared with genetic algorithms. The selection of these two algorithms is due to the characteristics to be able to model the event log correctly and can handle the control-flow. The capability in handling control-flow including the ability to detect hidden task, looping, duplicate task, detecting implicit/explicit concurrency, non-free-choice, the ability to mine and exploiting time, overcoming noise, and overcome incompleteness. The results of conformance checking on the heuristic miner algorithm for all data, fitness values, position, and structure are 1, 0.495, and 1, while the results of the genetic algorithm are 0.977, 0.706 and 1. Both algorithms have good ability in modeling processes and have high accuracy. The results of the F-score calculation on the heuristic miner algorithm for all data is 0.622, while the result in the genetic algorithm is 0.820. It indicates that genetic algorithms have better performance in modeling event logs based on process cube.