一种异构集成学习的儿科疾病诊断方法研究

A METHOD OF PEDIATRIC DISEASE DIAGNOSIS BASED ON HETEROGENEOUS INTEGRATION LEARNING

为提升儿科医生诊断效率及准确性,采取一种基于数据挖掘与机器学习相结合的分类技术。通过收集一定量的儿科患者病历样本,参考儿科医学书籍及资料,并利用数据挖掘的方法,从这些样本中提取出患者的病症和疾病数据,建立机器学习算法模型。对这些模型结果采用融合的方法进行集成,从而预测出未知样本的疾病结果。医生等相关人员可以利用这些结果作为参考,进而得出疾病诊断的结论。通过对总体样本"自助法"划分训练集与测试集,集成并预测,得到的实验结果表明,集成模型与其构成的单一模型相比,准确率提高了约6%。

To improve pediatrician diagnostic efficiency and accuracy,a classification technique based on data mining and machine learning is adopted. By collecting a certain amount of pediatric patients＇ medical records,referring to pediatric medical books and information,and using data mining methods,the patient＇s illness and disease data were extracted from these samples to build a machine learning algorithm model. We integrated these model results by means of fusion to predict the disease outcome of unknown samples. Doctors and other relevant personnel used these results as a reference,and then come to the conclusion of disease diagnosis. By dividing the training set and the test set for the overall sample ＂self-help law ＂,the experimental results obtained show that the accuracy of the integrated model is improved by about 6% compared with that of the single model.