| At present, the diagnosis of disease is still in the stage of the traditional experience. The correctness and errors of diagnostic result are closely related to the level of doctors. Especially for those complex cases which are difficult to diagnose, people often turn to medical experts. However, the number of experts is limited, especially for small and medium hospitals, they have not obtained the conditions to set up experts in every clinic. This leads to a higher rate of misdiagnosis. Therefore combining advanced computer knowledge with medical information database to develop assistant diagnostic system, has become an important developing direction.Based on the study of the decision tree, this article introduces boosting integration and improved principal component analysis feature selection technology, fully takes into account the intelligibility of diagnostic result, as well as the generalization,stability and accuracy of diagnostic model. Then AD-BC4.5-PCAFS assistant diagnostic algorithm is presented based on improved PCA feature selection and C4.5 boosting integration technology. This method is divided into three main stages: the first stage, using PCAFS to select features, thus can simplify the establishment of classifiers and reduce the influence of redundant attributes; the second stage, using boosting integration technology which is based on C4.5 to establish a number of classification model; the third stage, taking a vote to new samples according to trained models and their weights. In this paper, the experiments are based on many real medical data sets. The results show that PCAFS feature selection method is completely correct, and it is more efficient and easier to understand than PCA features compression method; AD-BC4.5-PCAFS assistant diagnostic algorithm is more efficient than other classical methods. Finally, this article applies AD-BC4.5-PCAFS method to consultation system which is similar to assistant diagnostic system, and has achieved good effect. |
PDF Full Text Request