كاربرد شبكه هاي عصبي مصنوعي در پيش بيني حاملگي ناخواسته و مقايسه آنها با روشهاي مرسوم آماري

In all of the sciences, statistical modeling has been used for determination of relevances between variables and prediction. Model selection is dependent on the nature of variables and its underlying preliminary assumptions. New methods for model building are artificial Neural Networks. Because of their intelligent structure and flexibility, they are competitors for traditional statistical modeling. Neural networks are advanced in theory and application. We have introduced artificial neural networks and used them in order to predict unwanted pregnancy and showed that they are capable of more accurate predictions as compared to other model building methods by ROC analysis.By using womanʹs age, number of live daughters and sons as predictors and based on the ROC analysis, we have computed an area under ROC curves for logistic, probate regression, discriminant analysis model, 3:2:1 and 3:3:1 percepterons; 0.758, 0.757, 0.630, 0.768, 0.823 respectively. It seems that prediction of unwanted pregnancy using artificial neural networks are more accurate than traditional statistical models.