In recent years, the Taiwan government has been calling for the use of public transportation and has been popularizing pollution-reducing green vehicles. Passenger transport operators are being encouraged to replace traditional buses with electric buses, to increase their use in urban transportation. Reduced energy consumption and operating costs are important operational benefits for passenger transport operators, and driving behavior has a significant impact on fuel consumption. Although many literatures or real-world systems have addressed the issues related to reducing energy consumption with electric buses, these works do not involve the records collected from an on-vehicle battery management system (BMS). Accordingly, the results of analyses of existing works lack in-depth discussions, and therefore the applicability of existing works is insignificant. Therefore, in this study, driving data were collected using a battery management system (BMS), and vehicular power consumption was classified according to energy efficiency. Then, decision trees and random forest were applied to construct energy consumption analytical models. Finally, the driving behaviors that influence energy consumption were investigated. A case study was conducted in which a Taichung passenger transport operator’s electric bus driving data on urban routes were collected to construct energy consumption analytical models. The data consisted of two parts, i.e., vehicle records and route records. On the basis of these records, we considered the practicability and applicability of the analytical models by transforming the unstructured records into raw data. Passenger transport operators and drivers can leverage the obtained eco-driving indicators for different bus routes for energy savings and carbon reduction.

In recent years, most Explainable Artificial Intelligence methods have primarily focused on image classification. Although research on interpretability in image segmentation has been increasing, it remains relatively limited. As an extension of Grad-CAM, several methods have been proposed and applied to image segmentation with the aim of enhancing existing techniques and adapting their properties. However, in this study, we highlight a common issue with gradient-based methods when generating visual explanations—these methods tend to emphasize background information, resulting in significant noise, especially when dealing with image segmentation tasks involving complex or cluttered backgrounds. Inspired by the widely used Eigen-CAM method, this study proposes a novel explainability approach tailored for semantic segmentation. By integrating gradient information and introducing a sign correction strategy, our method enhances spatial localization and reduces background noise, particularly in complex scenes. Through empirical studies, we compare our method with several representative methods, employing multiple evaluation metrics to quantify explainability and validate the advantages of our method. Overall, this study advances explainability methods for convolutional neural networks in semantic segmentation. Our approach not only preserves localized attention but also offers a simpler and more intuitive CAM, which has the potential to play a crucial role in sensitive application scenarios, fostering the development of trustworthy AI models.

Social insurance plays crucial role for protecting the social functionality. Since different occupation would lead to different level salary and different risk, the social insurance for the people is usually designed by the characteristic of their occupation. In Taiwan, Ministry of Health and Welfare provides health care to citizens and guarantees the basic income of their elderly life. Accordingly, Ministry of Health and Welfare operates and sponsor many types of social insurances such as National Pension Insurance, Labor Pension Insurance, etc. However, due to the Population ageing, many social insurances are going to encounter the crisis of pension bankruptcy. Unfortunately, the traditional actuarial methods are based on many hypotheses to calculate cash flow, which cannot predict individuals' personal payment behavior. Besides, most traditional actuarial methods only can estimate short-term cash flow. As the result, the error of traditional actuarial methods would significantly be increased while the period of prediction target is increased. In order to predicting individuals' personal payment behavior with long-term period, we propose a Deep-Learning framework to build a prediction model. Meanwhile, the Recurrent Neural Network (RNN) architecture is adopted to effectively abstract individuals' payment behavior and precisely predict their future payment behavior with a long-term period. We conduct a comprehensive experimental study based on a real dataset collected through Taiwan's Ministry of Health and Welfare. The results demonstrate a significantly improved accuracy of our Deep Learning approach compared with the existing works.

Sensory navigation device is an important trend in the field of machine learning and data science. Nowadays, more and more sensory navigation devices are built for blind people. The core of such sensory navigation devices for blind people usually is implemented by an Image Recognition Method. To build an image recognition model, many tools and online machine learning platforms are proposed. However, these tools or platforms are not able to completely satisfy the requirements for sensory navigation device. To build a sensory navigation device with satisfying requirements for blind people, an ability of reducing the cost of model training and a capability of user-centric image recognition are the two main issues. Therefore, to address the above issues, we propose a novel approach, namely, DLSNF (Deep-Learning-based Sensory Navigation Framework). Our proposed DLSNF is built based on the YOLO architecture to deal with the reducing cost of model training and NVIDIA Jetson TX2 to take the user-centric image recognition into account. Based on our proposed DLSNF, the real-time image recognition can be trained well and conduct a sensory navigation to help blind people. At the same time, the train model is embedded in NVIDIA Jetson TX2 which is the fastest, most power-efficient embedded AI computing device. For the experiments, we evaluated our proposed DLSNF with a real-world dataset consisting of 4,570 images collected by part-time workers. The extensive experimental results show that our proposed DLSNF more effectively and efficiently beyond the existing baselines.

Since the concept of Industry 4.0 emerged, an increasing number of major manufacturers have incorporated relevant technologies to monitor machinery and schedule processes so as to increase yield and optimize production. However, most machinery monitoring technologies are far too expensive for small- and medium-sized enterprises. Furthermore, the production processes at small- and medium-sized enterprises are simpler and can thus be optimized without excessively complex scheduling systems. This study therefore proposed the use of cheaper add-on sensors for monitoring machinery and integrated them with an algorithm that can more swiftly produce results that meet multiple objectives. The proposed algorithm is meant to extend the capabilities of small- and medium-sized enterprises in monitoring machinery and scheduling processes, thereby enabling them to contend on an equal footing with larger competitors. Finally, we performed an experiment at an actual spring enterprise to demonstrate the validity of the proposed algorithm.