Article Preview
Top1. Introduction
Process mining (W. van der Aalst, 2016) is an emerging scientific discipline that allows for extracting knowledge from event logs, i.e., collections of historical execution data, available in modern business information systems. Process mining is primarily used to discover, monitor, and improve processes by applying various techniques to event logs generated by the execution of processes (W. van der Aalst et al, 2012). Process discovery, conformance checking, and process enhancement form the three main tasks in process mining, which have been extensively applied to business processes in numerous application fields, such as finance, logistics, and health care. In the literature, very few authors report on the application of process mining in production processes (Demtroeder et Al, 2019) (Lechner, 2020)., e.g., when compared to other application domains such as sales, procurement, banking, and insurance. In addition, existing works mostly focus on use cases at an abstract level (Reinkemeyer, 2020)
Various techniques have been proposed to extract these shapes from performance effects. However, most existing technologies only remove simple pattern or a complex pattern that is limited to a specific set of manually selected events. Recent works has shown that patterns can be identified as common languages (Ammons et Al, 2002). In this way, patterns can be viewed normally or in automatic restricted mode and pattern mining can be identified as a language learning problem.
The current approaches are essentially the same. Each uses dynamic effects or metafiles as input to the operating system and creates one or more common internal languages that represent the pattern or workflow. However, individual solutions differ from the main routes.
In this paper, we introduce an entirely new approach to the extraction pattern that addresses many of the limitations of current technologies. Our understanding is two-sided. First of all, we see that smaller patterns can be formed parallel to larger patterns. Second, we also note that the micro-model configuration can be parallel. Then we use this knowledge to divide our work into two parts. First we use a technique to extract two kinds of small patterns and configure them with standard algorithms to control the finite state automata and some special rules used by M. Gabel and Z.Su (Gabel et Al, 2008), the extraction is also performed Can be extracted by the token algorithm (Gabel & Su,., 2008).
In the latter case, we use the Hadoop MapReduce framework to extract and compose models. These patterns were represented as finite state automata or their regular expressions. In this section, we extract the small patterns using the same token extraction algorithm, but in parallel with the map stage, and compute these small patterns in a larger pattern in parallel as a reduce step. Our job is to improve the technologies available in the extraction patterns and parallel distribution process.
Our approach is implemented in the Java programming language with two log files for two programs. The size of the first log file is 20 GB and that of the second file is 30 GB, which is created by the manufacturer of the log file. We tested our approach in three clusters in the cloud and merged it with the first group of five machines, the second group of ten machines, and the third group of 20 other machines. The effects of our programs include call, answer, message ... etc.