Ram Maheshwari Logo Image
Polina Tanasevich

Process Mining 2024

Process Mining (PM) and Machine Learning (ML) to analyze log errors

Project Image

Abstract

Analyzing technical log data is essential for understanding user experience and identifying errors in employee interactions with applications. This process helps uncover bottlenecks impacting productivity and customer satisfaction.

Objectives:
1. Error Identification: A binary classification model was developed to detect application errors along the customer journey.
2. Path Optimization: PM was used to detect inefficiencies in the employee pathway, enhancing interaction with the application.
3. Efficiency Improvement: Eliminating errors to create a smoother and more productive user experience.

Process Stages:
1. Data Collection: Identifying log sources (servers, databases, client applications).
2. Data Processing and Aggregation: Filtering and organizing logs for analysis.
3. Data Analysis: Using PM and ML for error pattern detection.
4. Visualization: Creating dashboards to present key findings.
5. Implementation: Developing error-resolution plans and monitoring improvements.
6. Feedback and Updates: Gathering user feedback to enhance the process.

Hypotheses and Metrics: Several hypotheses were tested, such as the effect of high load times on error frequency, alongside metrics like Bounce Rate and Error Rate.

Process Mining Techniques Used:
1. Process Discovery: Mapping user interactions for error-prone steps.
2. Performance Analysis: Identifying delay points.
3. Clustering: Grouping sessions by behavior patterns.
4. Anomaly Detection: Identifying unusual events for focused attention.

Tools Used

Python
Natasha
SberPM
Keras
NLTK
Gensim
BI-systems
SQL
Hadoop
Git
Scientific Research