基于多任务支持向量机的健康数据融合方法

针对电子健康服务管理中的多源数据融合难题,利用人工智能技术,结合多任务学习理论与支持向量机理论提出基于多任务支持向量机的数据融合方法(multi-task support vector machine for data fusion,简称mSVMDF).该方法对具有相同数据源的特征向量构造基于支持向量机的融合模型,在多任务学习框架下考虑结构稀疏性与各模型关联性的有机结合,以实现对具有不同数据源个数的多源数据的融合,并以多源影像数据与常规检验数据融合为例,开展数值实验验证方法的有效性.实验结果表明mSVMDF方法可以有效地融合具有不同数据源个数的多源数据,同时该方法具有较好的分类性能与结构稀疏性.

Health data fusion method based on multi-task support vector machine

Aiming at the difficult problems of multi-source data fusion in e-health service management, a multi-task support vector machine for data fusion (mSVMDF) method, which combines multi-task learning theory and support vector machine theory from the perspective of artificial intelligence, is proposed. In mSVMDF method, the data fusion model is obtained by the feature vector having the same type data source based on support vector machine theory, and both structure sparsity and tasks dependency is considered under the multi-task learning framework, to achieve multi-source data fusion with different number of data sources. Extensive multisource images and examination data fusion experiments are conducted to reveal the performance of mSVMDF method. The experimental results show that the proposed method can effectively overcome the difficulties of multi-source data fusion with different number of data sources and has good classification performance and structure sparsity.