Prediction of Pediatric Risk Using a Hybrid Model Based on Soft Computing Techniques

Abstract

We present an automatic system for the prediction of mortality risk in pediatric patients, which uses Soft Computing techniques instead of traditional ones based on score. The hybrid model applied combines both Case-Based Reasoning and Artificial Neural Networks with fuzzy set theory, taking its applications the advantages of these approaches. While the new way of prediction, named SAPRIM (Automated Predictor System of Infant Mortality Risk), was automatically defined from domain examples reducing the knowledge engineering effort, the experimental results using cross validation showed good accuracy with respect to other traditional classifiers. Besides, SAPRIM allows a more natural framework to include expert knowledge by using linguistic terms. After this automatic system was exploited by human experts for a year, the field evaluation corroborates good results.