Application of the gray model GM (1, 1) in predicting birth defects / 西安交通大学学报(医学版)

ABSTRACT

Objective: To investigate the application of gray model GM(1, 1) in predicting the incidence of birth defects at different levels and the effect of data volatility on the prediction outcome. Methods: Based on the monitoring data of birth defects in Xi'an from October 2009 to September 2016, the GM(1, 1) was used to predict the overall incidence of birth defects and incidence of five main birth defects at three levels (month, quarter, and year). We compared the fitting accuracy of different level prediction models. Results: The average relative error for yearly prediction of overall birth defect was 4.6%, and the mean square deviation was 0.259, which might suggest better prediction. Quarterly forecasting results were almost qualified and the average relative error was 10.2%. Monthly prediction was poor with an average relative error of 17.5%. With the extension of the forecast period, the grey model prediction results of the top five birth defects (congenital heart disease, cleft lip and palate, neural tube defects, multiple fingers, and congenital hydrocephalus) in Xi'an all increased, and the fitting accuracy gradually improved. The gray scale of the year was the best. Conclusion: The prediction results of the gray model may be related to the volatility of the data. It may be suitable for predicting the incidence of birth defects by the year.