An Artificial intelligence technique for the prediction of persistent asthma in children

The prediction of asthma that persists throughout childhood and into adulthood, in early life of a child has practical, clinical and prognostic implications and sets the basis for the future prevention. Artificial Neural Networks (ANNs) seems to be a superior tool for analyzing data sets where nonlinear relationships are existing between the input data and the predicted output. This study presents an effective machine-learning approach based on Multi-Layer Perceptron (MLP) neural networks, for the prediction of persistent asthma in children. Through a feature reduction, 10 high importance prognostic factors correlated to persistent asthma have been discovered. The feature selection approach results in 89.8% reduction of the initial number of features. Afterwards, a feature reduced classifier is constructed, which achieves 100% accuracy on the training and test data sets. Experimental results are presenting and verify this statement.