摘要
针对精神病的致病因素较复杂、缺乏有效风险预测方法,导致发病率越来越高、患者越年轻化的问题,提出一种基于深度学习的精神病风险预测方法。方法研究深度协同过滤算法对海量特征数据进行预处理,建立患者特征和健康隐式特征间的高阶非线性交互关系模型,利用面向隐式反馈信息算法,估算健康隐式特征间的相似度;通过研究反向传播算法计算特征向量的置信度,从训练数据集中学习健康隐式特征,构建精神病风险预测目标标签和评价指标体系,以实现更高效、准确的个性化的精神病风险预测算法。实验结果表明所提算法比传统算法有更好的性能。
Aiming at the problem that the incidence of psychosis risk is getting higher and higher, patients are getting younger and younger, and there is no effective prediction method. This paper proposes a psychosis risk prediction method based on deep learning. With this method, the depth collaborative filtering algorithm was used to preprocess the data, estimate the missing data and add it back to the data set. At the same time, the depth confidence network was used to process the complex multimodal data, so as to achieve the effect of psychosis risk prediction. Experimental results show that the proposed algorithm is better than the conventional deep confidence network and machine learning methods.
作者
沈贝敏
周小平
孙卫国
叶韶光
SHEN Bei-ming;ZHOU Xiao-ping;SUN Wei-guo;YE Shao-guang(Shanghai Mental Health Center,Shanghai 200030,China;Shanghai Normal University,Shanghai 200234,China)
出处
《计算机仿真》
北大核心
2020年第10期417-420,共4页
Computer Simulation
基金
上海市自然基金(16ZR1424500)。