Anomaly Detection In Process Mining

This is the repository for "Anomaly detection in business process event logs: a comparative study between neural networks-based and count-based approaches"

Folder estructures

├── anomaly_detection 
├── research_questions
│   ├── best_parameters 
│   ├── ├── logs 
│   ├── ├── resultados
│   ├── best_anomaly_detection_and_descovery 
│   ├── ├── logs 
│   ├── ├── resultados
├── LICENSE 
└── README.md 

The main folders are:

• anomaly_detection folder containing all code
• research_questions. At the beginning, this folder only contains logs. After execution, this folder will contain results of each phase of research

Guide to use

Steps

1. Install Anaconda

2. Install R (Some R statistics functions are necessary for this python code)

3. Create a conda environment using env_pyth3.8.yml ( made available in repository)

   conda env create -f env_pyth3.8.yml

   conda activate env_pyth3.8

4. Install pm4py following instructions in https://pm4py.fit.fraunhofer.de/install-page

   1. Install Microsoft Visual C++ Redistributable
   2. Install graphviz 3.0.0
   3. Install Pm4py 1.3.2

   conda install pm4py==1.3.2

5. Execute jupyter

   jupyter notebook notebooks

6. Execute notebook

Phases

We have three main phases in the research:

1. Look for the best parameterization
2. Analyze robustness of approaches
3. Analyze approaches to process discovery improvement

Automatizar.py

The automatizar.py file contains the main code. In order to execute phase 02 in main.py we can set as follow:

phase_number = 2

In order to execute phase 03 in main.py we can set as follow:

phase_number = 3

main.py was set using nome_grupo_experimentos = "test_case". In order to get the same results as in dissertation, nome_grupo_experimentos = "relatorio1-question2" should be used.

Notebooks

Notebook01:

It executes phase02 and save results in folder "anomaly_detection\research_questions\best_anomaly_detection_and_descovery\resultados"

• Results of experiment group [id_group] will be saved in "\resultados\resultados_gexp_binetn_[id_group]\gexp_binetn_[id_group]_all_results_details.csv".
• A report in .csv will be saved in gexp_binetn_[id_group]_all_results_details.csv. That file include all information stored about experiments

Notebook02:

It executes phase03 and save results in folder "\resultados_gexp_binetn_[id_group]\logs_postproces_batch1"

• Filtered logs will be saved in \logs_postproces_batch1\filtered_logs
• Discovered models will be saved in \logs_postproces_batch1\petrinets
• Report about quality of discovered models will be saved in \logs_postproces_batch1\gexp_binetn_[id_group]_all_results_details_dscvry_batch1.csv