Skip to content

Latest commit

 

History

History
121 lines (85 loc) · 3.59 KB

process_mining_guide.md

File metadata and controls

121 lines (85 loc) · 3.59 KB

Process Mining Guide

Version: v1.7.0 Status: 🟢 Production-Ready Last Updated: 2026-03-09

Overview

ThemisDB's Process Mining module enables automated discovery and analysis of business processes from event log data. It provides a comprehensive suite of algorithms for process discovery, conformance checking, and performance enhancement.

Key Features

Process Discovery Algorithms

  • Alpha Miner: Classical algorithm for structured processes
  • Heuristic Miner: Robust algorithm handling noise and exceptions
  • Inductive Miner: Sound process models with guaranteed fitness

Analysis Capabilities

  • Directly-Follows Graph (DFG): Visualize process flows with frequencies
  • Variant Analysis: Identify common and rare process execution patterns
  • Bottleneck Detection: Find performance issues and delays via performance enhancement

Conformance Checking

  • Token replay for fitness calculation (fast; measures how well the model explains the log)
  • Alignment-based conformance (precise, but slower; finds minimal-cost edit sequence between model and log)
  • Deviation detection and analysis

Export Formats

  • BPMN 2.0 XML for standard process modeling
  • Petri Net (PNML) for formal analysis
  • ThemisDB process definition (saved to _process_definitions)

Usage Examples

Basic Process Discovery

ProcessMining mining(db);  // requires RocksDBWrapper&

// Extract event log from collection
EventLogConfig config;
config.case_id_field   = "order_id";
config.activity_field  = "action";
config.timestamp_field = "timestamp";

auto [status1, event_log] = mining.extractEventLog("audit_log", config);
if (!status1.ok) { /* handle error */ }

// Discover process model (MiningConfig wraps MiningAlgorithm)
MiningConfig mcfg;
mcfg.algorithm = MiningAlgorithm::HEURISTIC;

auto [status2, model] = mining.discoverProcess(event_log, mcfg);

// Export to BPMN
auto [status3, bpmn] = mining.exportToBPMN(model);

Variant Analysis

auto [status, variants] = mining.analyzeVariants(event_log);

// Find most common process variant
if (!variants.empty()) {
    const auto& most_common = variants[0]; // sorted by frequency desc
    std::cout << "Most common variant occurs "
              << most_common.frequency << " times ("
              << most_common.percentage << "%)" << std::endl;
}

Bottleneck Detection

Bottleneck detection requires a performance-enhanced process model:

// First enhance model with performance data
auto [s1, enhanced] = mining.enhanceWithPerformance(model, event_log);

// Then detect bottlenecks (returns names of bottleneck nodes)
auto [s2, bottleneck_nodes] = mining.detectBottlenecks(enhanced, 0.9); // 90th percentile

for (const auto& node_name : bottleneck_nodes) {
    std::cout << "Bottleneck node: " << node_name << std::endl;
    // Performance details are in enhanced.node_avg_duration[node_name]
}

Conformance Checking

auto [s, conformance] = mining.checkConformance(event_log, model);
std::cout << "Fitness: " << conformance.fitness << "\n";
for (const auto& dev : conformance.deviations)
    std::cout << "Deviation: " << dev << "\n";

Related Documentation

Last Updated: 2026-03-09 Version: v1.7.0 Status: 🟢 Production-Ready