Multi-class diagnosis classification on high dimension data by logistic models

Logistic regression has been increasingly used in chronic gastritis research. Using expression of logistic regression monitored simultaneously by maximum likelihood estimation, contribution of gastritis symptom to the syndrome classifications are distinguished, and chronic gastritis samples are more accurately classified. While logistic regression has been extensively evaluated for dichotomous classification, there are only limited reports on the important issue of multi-class chronic gastritis classification. It needs to explore the logistic regression of the multi-class chronic gastritis classification. In this research, we address multi-class chronic gastritis classification by applying logistic regression based methods on data of nominal and ordinal scaled sample class outcomes, e.g., samples of different chronic gastritis subtypes. Logistic regression based classifiers are assessed by accurate classification rates on chronic gastritis data and comparing with HGC model discrimination based classifiers. The result shows that classify performance derive from logistic regression model has the advantage over traditional model by 26.94%.