Fall is a major threat to stroke survivors with the problems of gait and balance disorders in the rehabilitation phase following severe consequences on quality of life and a heavy burden to their families. Many solutions have been proposed to assess fall risk for elders based on inertial sensor-based signals, however, there still exists a great challenge of transferring them from elderly populations to the stroke-survivors populations as gait disorder patterns are significant difference between elders and stroke survivors. In this study, we conduct a pilot study to collect inertial sensor-based signals from stroke survivors when they performed the timed up and go test, and build an automatic fall risk assessment model with the architecture of Siamese network, with a merit of mitigating the problem of small sample size. Specifically, the proposed automatic fall risk assessment model consists of two parallel convolutional neural networks, each of which is composed of three convolutional layers, two max-pooling layers, and three fully connected layers. To utilize the space relation among accelerator-based and gyroscope-based signals, two-dimensional discrete wavelet transform extracts image-like features, wavelet coefficients, from inertial sensor-based signals as the input. Experimental results show that the proposed fall risk assessment model has achieved a promising results, which outperform cutting-edge methods with a big margin. The proposed fall risk assessment model with low computational complexity and limited memory consuming can be deployed on an embedded system to provide fall risk assessment service for stroke survivors in point-of-care environments or community settings.

Influenza outbreaks have brought increasing challenges to public health systems globally. The effective and efficient tracking of influenza can help authorities make informed and proactive decisions. In this study, we focus on nowcasting influenza epidemics at regional level. To alleviate the information lag between the release of the Centers for Disease Control's influenza-like illness (ILI) reports and real-time influenza activity, we incorporate Google search data and holiday effects to facilitate the ILI nowcasting. We develop a deep learning framework by extending the spatiotemporal residual network (ST-ResNet) to nowcast ILI rates in irregular-shaped region. We investigate the effect of temporal and spatial dependencies among irregular-shaped regions as well as external influence on ILI nowcasting. Various forecasting models, including time series models, penalized regression models, and other deep learning models, are employed for evaluating the performance of the proposed framework. Moreover, based on city-level ILI data in Taiwan, we conduct extensive experiments for methods validation. The results show that the extended ST-ResNet can effectively capture the complex spatiotemporal dependencies of ILI activity and the effects of external variables. Additionally, the strategy of incorporating Google search data and holiday effects improves the prediction. The findings may provide insights into the utility of various statistical and deep learning methods for effective ILI tracking at regional level and facilitate informed decision making for public health authorities.

The queries entered into search engines register hundreds of millions of different searches by tourists, not only reflecting the trends of the searchers' preferences for travel products, but also offering a forecasting of their future travel behavior. This paper proposed a forecasting framework based on internet search index and machine learning to forecast tourist arrivals, and compared the forecasting performance of two different search engines data, Baidu and Google. The empirical results suggest that the proposed KELM models by fusing Baidu index and Google index can significantly improve the forecasting performance and outperform other benchmark models in terms of forecasting accuracy.