An Improved Framework of Tree-Structured Data Mining for Business Process Logs Analysis

An Improved Framework of Tree-Structured Data Mining for Business Process Logs Analysis

• 1.
  An Improved Frameworkof Tree-Structured Data Mining for Business Process Logs Analysis Izwan Nizal Mohd Shaharanee Jastini Mohd Jamil UNIVERSITI UTARA MALAYSIA
• 2.
  Introduction Business process | organized| measured arrangement | of activities Complex business process required an automated and systematic process management Business Process Model (BPM) is a method, techniques and tools that support the design, management and analysis of operational business process. BPM were supported in e- Mail applications, inventory software, account system ect.
• 3.
  Introduction Business process are import for process optimization and improvement Examples: University Database, Business Workflowsand many more Knowledge from these processes can be a great insight for better decision making process Instruction detection, audit trails, crime surveillance, fraud transaction are several potential area for business process mining
• 4.
  RELATED WORKS Business processleaves traces and event logs Give valuable information about the process captured Process discovery task to discover best workflow for business process Log data stored in XML format(MXML/XES) The frequent subtree mining methods are the basis for discovering interesting associations among tree-structured data objects in XML data
• 5.
  RELATED WORKS The utilizationof subtree mining in process log data is less explored A subtree would capture the workflow execution pattern over the business process The subtree patterns also preserve the context of the events and event attributes within a trace
• 6.
  Problem Definitions • Representednot in user-friendly format. Large Number of Event Logs/Business Process Data: • Large and complex Repetitive Events • Conventional Storing ApproachHidden Knowledge
• 7.
  Problem Statement Event logsand business process are often very large in terms of the number of records and they are usually represented in a format that is not user-friendly. Hence, sophisticated, robust methods and tools are needed to extract hidden knowledge from event data to optimize or improve business processes
• 8.
  The Objective In thisresearch paper, a direct mining approach for XML-based process logs were investigated and an improved framework is proposed
• 9.
  The RESEARCH FRAMEWORK
• 10.
  The PHASES Pre-Processing DSM Extraction KnowledgeDiscovery Intepretation
• 11.
  Phase 1: Pre-Processing Raw ProcessData ETL (Extract, Transform, Load) XES/MML Grouping, Discretization, Filtering Transformation
• 12.
  Phase II : DSMExtraction DSM(Documents Structure Model) FDT(Flat Data Table)
• 13.
  DSM – transformingtree data into flat representation Extract the most general tree structure (every instances matches) Become the first rows Remaining instances is placed based on label, backtrack and non- existance
• 14.
  Phase III : KnowledgeDiscovery Knowledge Discovery based on FDT Frequent Subtree Mining Prediction Clustering
• 15.
  Phase IV : Interpretation MappingKnowledge Model to DSM Database Structure Model (DSM) Regenerate Subtree Interpretation
• 16.
  The CONCLUSION • Howthe proposed framework works ? Business Logs Data (Tree- Structured Format) Extract their Document Structure Model (DSM) Business Logs Data in Flat Data Format Mining the DataObtain Knowledge Mapping the Knowledge to DSM Interpret the Knowledge
• 17.
  The CONCLUSION • Businessprocess logs exist in many areas • The capabilities of representing process logs into XML format offer a better data analysis • In this paper XML–based business process logs were presented into flat data format • Thus, offer many classical data mining techniques to be utilized