摘要
针对目前帕金森病早期预测方法普遍存在误诊率高、步骤繁多等问题,设计了基于AdaBoost算法的按键动作识别方法,实现对帕金森病早期的精准预测。该方法首先删除数据集的缺失值,并选取按键次数过万的数据;然后针对不同按键手,根据按键的时间间隔对预处理后的结果进行分类,以平均值、标准差、方差、偏度和峰度5个指标为特征,对每一位病人的数据进行分块,扩充数据集,并加入高斯噪声平衡数据集;最后应用AdaBoost算法进行分类预测。在公开的数据集上进行实验,结果表明:在按键数据集分类上,该方法的准确率、灵敏度和特异性分别为95%、98%和97%。该方法具有较高的准确率、灵敏度和特异性,为帕金森病早期的精准预测提供了一种有效的解决方案。
Aiming at the problems of high misdiagnosis rate and numerous steps in the early prediction method of Parkinson′s disease, a keystroke recognition method based on AdaBoost algorithm was designed to achieve accurate prediction of early Parkinson′s disease. By adopting this method, the missing values of the dataset were deleted, and the data with more than 10 000 keystrokes were selected. Then, for different keystroke hands, the results after pre-treatment were classified according to the time interval of the keys, the five indicators of calculated mean, standard deviation, variance, skewness and kurtosis were used as new features, and the data of each patient was blocked, so as to expand the dataset and Gaussian noise was added to balance the dataset. Finally, the predictive classification was carried out by AdaBoost algorithm to achieve early prediction of Parkinson′s disease. Experiments were conducted on the published dataset, and the experimental results showed that the accuracy, sensitivity and specificity of the algorithm were 95%, 98% and 97%, respectively, in the classification of the key dataset. Compared with other algorithms, this method has high accuracy, sensitivity and specificity, providing an effective solution for early prediction of Parkinson′s disease.
作者
许昊
童基均
齐鹏嘉
周思薇
XU Hao;TONG Jijun;QI Pengjia;ZHOU Siwei(School of Information Science and Technology,Zhejiang Sci-Tech University,Hangzhou 310018,China;Zhejiang Rehabilitation Medical Center,Hangzhou 310014,China)
出处
《浙江理工大学学报（自然科学版）》
2023年第1期83-88,共6页
Journal of Zhejiang Sci-Tech University(Natural Sciences)
基金
浙江省医院协会管理软科学项目(2021ZHA-KEB206)
浙江省自然科学基金项目(LQ22F010006)。
