Comparison of BP Neural Network Model and Logistic Regression in the Analysis of Influencing Factors of Violence in Hospitals

Abstract

Collecting violence cases for medical personnel from different levels of the hospital of Tangshan, we create a model for influential factors of hospital violence, and respectively with BP Nerve Network and logistic regression, by sensitivity, specificity and ROC curve, it is compared with two methods,in order to discovering effective analytical method . The training set and testing set sensitivity of BP Neural Network Model are 0.916 and 0.935,and the specificity is 0.447 and 0.526,the area of ROC curve is 0.769 and 0.785；for logistic regression Model ,for its the training set and testing set, sensitivity is 0.907 and 0.925, the specificity is 0.377 and 0.404, the area of ROC curve is 0.663and0.666. In hospital violence influencing factors, the forecast capability of BP Neural Network Model is better than logistic regression Model and it has farther extend value.