Accuracy of intelligent medical systems

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Machine Learning for Healthcare

Background:

• Intelligent medical systems, a subset of medical software, must ensure accurate predictions.
• System accuracy hinges on factors like basic methods (decision trees, neural networks), induction methods (purity measures), and support methods (discretization, pruning, boosting).

Purpose of the Study:

• To investigate the influence of various methods on the accuracy of decision trees.
• To introduce and evaluate novel hybrid purity measures for enhanced decision tree induction.

Main Methods:

• Extensive research conducted on 54 UCI databases.
• Evaluation of different basic methods, induction methods (purity measures), and support methods (discretization, pruning, boosting).
• Development and testing of new hybrid purity measures.

Main Results:

• The selection of purity measures, discretization techniques, and boosting methods significantly impacts decision tree accuracy.
• Novel hybrid purity measures demonstrated superior performance over existing measures on certain datasets.
• Optimizing these factors leads to a notable increase in the accuracy of induced decision trees.

Conclusions:

• Careful selection and combination of purity measures, discretization, and boosting are crucial for improving intelligent medical system accuracy.
• The process of identifying optimal factors for decision tree accuracy is complex and resource-intensive.