【作者】 殷菲；

【导师】 潘晓平；

【作者基本信息】 四川大学 ， 流行病与卫生统计学， 2004， 硕士

【摘要】 新生儿黄疸预测模型 目的 现有的用于分类问题的方法的重要基础是传统统计学,前提是足够多样本,当样本数目有限时难以取得理想的效果。因此当传统分类方法用于新生儿黄疸预测问题时,未必能取得好的效果。支持向量机是一种新的分类学习算法。本研究应用支持向量机建立预测模型,对新生儿黄疸进行预测,同时也用概率神经网络方法建立预测模型对同样的数据进行预测,比较两种方法的预测效果。 方法 应用支持向量机理论建立新生儿黄疸诊断预测模型。为了防止发生过度拟合,在模型拟合过程中运用了交叉验证的方法。同时,建立基于概率神经网络的预测模型,并且对两种模型的预测结果进行比较。 结果 支持向量机的分类结果与实际情况吻合得较好。支持向量机的分类误差与概率神经网络相近,但支持向量机的使用较概率神经网络简单。 结论 支持向量机正在成为机器学习领域的一个新的热点,用支持向量机来解决分类问题是一种新的视角,有着光明的前景。更多还原

【Abstract】 Prediction Model for the Diagnosis of Neonatal Jaundice Objective: Most of the existing methods used in classification are based on traditional statistics, which provides conclusion only for the situation where sample size is tending to infinity. When confronted with the problem of predicting the neonatal jaundice, most exiting methods may not work well. In this paper, Support Vector Machine, which is a new technique for data classification, was used to establish a prediction model for neonatal jaundice. In addition, the method of Probabilistic Neural Network was also applied to this issue as a contrast.Method: A prediction model based on the Support Vector Machine theory was established for this issue, in which cross-validation procedure was applied to prevent the over-fitting problem. Besides, another model based on Probabilistic Neural Network was also established for this issue. Then, the generalization error of those two models were obtained and compared.Result: The classification results of the Support Vector Machine are in good accordance with the observed values. The generalization error of Support Vector Machine is approximately equal to that of Probabilistic Neural Network. And Support Vector Machine is easier to use than Neural Networks.Conclusion: It is believed that the study of SVM is becoming a new hotarea in the field of machine learning. Using Support Vector Machinetheory to solve classification problem is a method with promisingprospect.更多还原