Extraction frequent patterns in trauma dataset based on automatic generation of minimum support and feature weighting

Area of Science:

• Health Informatics
• Data Mining
• Medical Data Analysis

Background:

• Data mining techniques, particularly frequent pattern discovery, are crucial for analyzing health data to diagnose diseases and identify contributing factors.
• The Kashan Trauma Registry contains valuable health information that can benefit from advanced data mining methods.

Purpose of the Study:

• To discover frequent patterns within the Kashan Trauma Registry data using a novel data mining approach.
• To enhance the efficiency and effectiveness of frequent pattern discovery in healthcare datasets.

Main Methods:

• Utilized real-world data from the Kashan Trauma Registry.
• Applied the classical Apriori algorithm and a new method for frequent pattern and rule extraction.
• Developed a new method incorporating variable weighting and harmonic mean for automatic minimum support calculation in Python.

Main Results:

• The new method dynamically calculates minimum support, unlike the manual, single-value approach of the classical Apriori algorithm.
• Demonstrated superior performance of the new method over the classical Apriori algorithm in terms of memory consumption, execution time, and the number of patterns and rules discovered.

Conclusions:

• Manually setting minimum support in data mining is inefficient, increasing execution time and memory usage, especially for users unfamiliar with the dataset.
• The new method provides valuable insights into trauma data, highlighting the importance of elderly care, public education on accident response, and motorcycle safety.
• Frequent pattern discovery in trauma registries can inform public health strategies and improve patient outcomes.