ABDUCTIVE MACHINE LEARNING FOR MODELING AND PREDICTING THE EDUCATIONAL SCORE IN SCHOOL-HEALTH SURVEYS

The use of modern abductive machine learning techniques is described f or modeling and predicting outcome parameters in terms of input parame ters in medical survey data. The AIM(R) (Abductory Induction Mechanism ) abductive network machine-learning tool is used to model the educati onal score in a health survey of 2,720 Albanian primary school childre n. Data included the child's age, gender, vision, nourishment, parasit e infection, family size, parents' education, and educational score. M odels synthesized by training on just 100 cases predict the educationa l score output for the remaining 2,620 cases with 100% accuracy. Simpl e models represented as analytical functions highlight global relation ships and trends in the survey population. Models generated are quite robust, with no change in the basic model structure for a 10-fold incr ease in the size of the training set. Compared to other statistical an d neural network approaches, AIM provides faster and highly automated model synthesis, requiring little or no user intervention.