Search

During the past few years, severe air-pollution problem has garnered worldwide attention due to its effect on health and wellbeing of individuals. As a result, the analysis and prediction of air pollution has attracted a good deal of interest among researchers. The research areas include traditional machine learning, neural networks and deep learning. How to effectively and accurately predict air pollution becomes an important issue. In this paper, we propose an Aggregated LSTM (Long Short-Term Memory) model (ALSTM) based on the LSTM deep learning method. In this new model, we combine local air quality monitoring station, the station in nearby industrial areas, and the stations for external pollution sources. To improve prediction accuracy, we aggregate three LSTM models into a predictive model for early predictions based on external sources of pollution and information from nearby industrial air quality stations. We exploited the data with 17 attributes collected by Taiwan Environmental Protection Agency from 2012 to 2017 as the training data to build the ALSTM forecasting model, and we tested the model using the data collected in 2018. We conducted some experiments to compare our new ALSTM model with SVR (Support Vector Machine based Regression), GBTR (Gradient Boosted Tree Regression), LSTM, etc., in the prediction of PM2.5 for 1–8 h, and evaluated them using various assessment techniques, such as MAE, RMSE, and MAPE. The results reveal that the proposed aggregated model can effectively improve the accuracy of prediction.

In this paper, we proposed a method for accelerating the transmission of a series of streaming files for multi-session file delivery scenarios. When streaming files are transmitted, the throughput of data transmission can be increased and the out of order problem of file reception (i.e., file reception delay) can also be reduced at the same time. This is achieved by using a repair file generation mechanism based on AL-FEC, which is triggered while an empty file transmission session appears but there are still uncompleted files waiting to be delivered. Performance evaluation results shows that the file transmission throughput of the proposed method are 28% to 44% higher than the throughput of a traditional file transmission method that is designed for reducing file reception delay. However, the file reception delay is the same as the above-mentioned traditional method.

Due to flexible scheduling requirements for various applications, a cloud platform usually has some temporarily unleased workers. To make cost effective, such considerable amount of idle workers can be collected to perform malleable tasks. These workers, called spot workers, often cannot work unstably since they may be interrupted by an external scheduler. This paper proposes a resource management approach that employs replication to increase resource availability. In fact, lifting the replication factor can increase the reliability that a worker can finish a job. But this oppositely decreases resource-usage efficiency due to computational redundancy. We also develop an algorithm that not only controls the replication factor to adapt the changing workload, but also maintains the system performance.

As advance of economy and industry, the impact of air pollution has gradually gained attention. In order to predict air quality, there were many studies that exploited various machine learning techniques to build predictive model for pollutant concentration or air quality prediction. However, enhancing the prediction performance always is the common problem of existing studies. Traditional templates based on machine learning and deep learning methods, such as GBTR (gradient boosted tree regression), SVR (support vector machine-based regression), and LSTM (long short-term memory), are most promising approaches to address these problems. Some previous researches showed that ensemble learning technology can improve predictive performance of other domains. In order to improve the accuracy of forecasting, in this paper, we propose a hybrid model and framework to improve the forecasting accuracy of air pollution. We not only exploit stacking-based ensemble learning scheme with Pearson correlation coefficient to calculate the correlation between different machine learning models to integrate various forecasting models together, but also construct a framework based on Spark+Hadoop machine learning and TensorFlow deep learning framework to physically integrate these models to demonstrate the next 1 to 8 h’ air pollution forecasting. We also conduct experiments and compare the result with GBTR, SVR, LSTM, and LSTM2 (version 2) models to demonstrate the proposed hybrid model’s predictive performance. The experimental results show that the hybrid model is superior to the existing models used for predicting air pollution.

ONVIF (Open Network Video Interface Forum) is an importance industrial standard for the video surveillance field. There are over 8000 ONVIF-compliant IP cameras (i.e., ONVIF NVT) and network video recorders (i.e., ONVIF NVS) from various vendors listed on the ONVIF official website. However, because the current ONVIF specifications do not support any fault-tolerant functions to handle NVS failure, in this paper we propose a fault-tolerant solution for ONVIF NVS and try to minimize the modifications to the current ONVIF specifications in the proposed design. Besides, we also propose an extension to the ONVIF NVT for preventing the loss of captured NVT video data during the failover procedure triggered by NVS failure. Finally, we evaluate the following four performance metrics of the proposed fault-tolerant ONVIF surveillance system: the maximal video local cache time supported by an NVT, the response time for accessing an ONVIF web service, the interrupt duration of video recording due to NVS failure, and the live video black out time observed by a video management system. Since the experimental results are all within the acceptable range for a real fault-tolerant video surveillance system in the real world, the experiences provided by this paper is a good reference for improving the fault-tolerant capabilities of the ONVIF video surveillance systems.

Futures market has high leverage and the characteristics of over-performing returns. Hence, it always attracts lots of investors. However, three major kinds of institutional traders are more influential than individual investors in Taiwan. In this paper, a monthly predicting model for the weighted price index of the Taiwan Stock Exchange (TAIEX) was built based on correlation and regression analysis by using the following parameters: Dow Jones industrial index, NASDAQ index, M1A, M1B, M2 annual growth rate, US dollar exchange rate, economic monitoring indicator and global oil prices. Then, based on the prediction results, we define two trading strategies and apply them in trading MTX (mini Taiwan index futures). The evaluation result shows that both the two trading strategies have good returns.

Due to the growing popularity of Internet of People (IoP) and its impacts on human activity, it has quickly become an important research field and hot subject. Since one purpose of IoP is to connect people to people, personal live video delivery that has been popular recently can also be regarded as an important application of IoP. For improving the video quality of video communication services, Raptor code has been adopted in various broadband communication systems. In this paper, we propose a systematic Raptor code decoder based on the group parameters for a group of encoding symbols. For a Raptor code application that can frequently use one or several fixed source block lengths (i.e., the number of source symbols in a source block), we could produce the corresponding group parameters in advance and use them to decode the received encoding symbols more efficiently. For personal live video delivery scenario based on the IPTV delivery, the simulation results show that our decoder is faster than the conventional Raptor code decoder which is adopted by the DVB and 3GPP specifications.

Raptor codes can provide good error correction capability and efficient encoding and decoding rate. Its fountain code property is effective in avoiding packet retransmission for both unicast and multicast service delivery. Hence, there are many mobile, vehicular and broadband applications adopting it such as mobile multimedia broadcasting, multimedia communications for high-speed rails and broadband IPTV systems. In this paper, we present an efficient systematic Raptor codes encoder based on operation lists for known source block lengths. For a Raptor codes application that can frequently use one or several fixed source block lengths (i.e., the number of source symbols in a source block), we could produce the corresponding operation lists in advance and use them to generate the encoding symbols more efficiently. We first introduce the basic architecture of the proposed Raptor encoder, and then describe the details about how to generate operation lists and Raptor intermediate symbols. The simulation results show that our encoder is at least two times faster than the conventional Raptor codes encoder which is adopted by 3GPP and DVB-H. Besides, the reduction on CPU utilization for a real Raptor-based streaming server is described, which is from 11.59 % to 53.41 %, depending on the employed source block length and symbol size.

Although cloud storage integration services have been developed in recent years that providing a single interface to manage multiple cloud storage accounts across different providers, these cloud storage integration services still remain several issues that do not address. First, because these services do not support arrangement mechanisms for users, they need to keep track of which account they want to upload files to and arrange their storage spaces manually. Second, since each account is divided into different sections, it is impossible to provide a virtual folder that contains files from different cloud storage providers. In addition, when all accounts are nearly full, integration services could not chunk a file to store it across multiple accounts. Third, uploading and moving files across accounts are always done through the integration service's server, which might become a bottleneck when there are a lot of users uploading files at the same time. In this paper, we propose a new cloud storage integration service - Unified Cloud Storage (UCS) for solving the above-mentioned issues, and evaluate its performance by experiments.

ONVIF (Open Network Video Interface Forum) is a popular standard for the video surveillance industries. Since ONVIF specifications do not provide fault-tolerant functions to deal with NVT (Network Video Transmitter) or NVS (Network Video Storage) failure, we propose a fault-tolerant solution for both ONVIF NVT and NVS in this paper. The proposed fault-tolerant extension is also possible to be deployed over cloud-based environments, such as IaaS (Infrastructure as a Service) platforms. The proposed solution is evaluated by the following three performance metrics: the access response time for ONVIF web services, the video recording interrupt duration due to NVS failure, and the video black out time of the surveillance video monitoring system. The experimental results show that all the metrics are within the acceptable range for a real world video surveillance system.