Using machine learning and expert systems to predict preterm delivery in pregnant women

Machine learning and statistical analysis were performed on 9,419 perinatal records with the goal of building a prototype expert system that would improve on the current accuracy rates achieved by manual pre-term labor and delivery risk scoring tools. Current manual scoring techniques have reported accuracy rates of 17-38%. The prototype expert system produced in this effort achieve overall accuracy rates of 53%-88% when tested on records that were not used in either statistical analysis or machine learning. Based on the success of this initial effort, the development of a full expert system to assist in pre-term delivery risk decision support, using the methods described in this paper, is planned