Search results

This chapter introduces a risk control framework on credit card fraud instead of providing a solely binary classifier model. The anomaly detection approach is adopted to identify fraud events as the outliers of the reconstruction error of a trained autoencoder (AE). The trained AE shows fitness and robustness on the normal transactions and heterogeneous behavior on fraud activities. The cost of false-positive normal transactions is controlled, and the loss of false-negative frauds can be evaluated by the thresholds from the percentiles of reconstruction error of trained AE on normal transactions. To align the risk assessment of the economic and financial situation, the risk manager can adjust the threshold to meet the risk control requirements. Using the 95th percentile as the threshold, the rate of wrongly detecting normal transactions is controlled at 5% and the true positive rate is 86%. For the 99th percentile threshold, the well-controlled false positive rate is around 1% and 83% for the truly detecting fraud activities. The performance of a false positive rate and the true positive rate is competitive with other supervised learning algorithms.

This study aims to examine whether global gold futures returns volatilities and trading activities are threshold cointegrated.

This study considers 11 gold futures markets, including 3 developed futures markets and 8 developing futures markets. This study also analyzes futures trading activities for speculators and hedgers. This study uses a nonlinear threshold vector error correction model (TVECM) and a threshold Lagrange multiplier (LM) test proposed by Hansen and Seo (2002).

The findings show that global gold futures return volatilities (FRV) and trading activities are not always threshold cointegrated. Most developed futures markets exhibit threshold cointegrated of gold FRV and trading activities for speculators and hedgers, whereas some developing futures markets exhibit threshold cointegrated. It suggests that speculators and hedgers trading activity conveys valuable information about changes in market volatility dynamics. On the other hand, responses to error-correction effect among gold FRV and trading activities for speculators and hedgers are dramatically different for developed and developing gold futures markets, respectively, particular in the unusual regime.

Research results show that threshold cointegration between global gold FRV and trading activities matters but not always. Thus, threshold relations have improved the authors’ understanding of global gold futures price discovery process with a threshold. For research limitations, this study uses only near month futures contracts, as it contains more information but not using far month contracts.

The findings may have important trading implications with additional insights in a(n) (un)usual regime further regulation may be detrimental to the price responsiveness in futures markets if increased price volatility and trading volume are attributed to liquid and efficient markets.

The findings may have important policy implications with additional insights. For example, in a(n) (un)usual regime greater regulatory restrictions may be warranted to decrease market inefficiencies if increased price fluctuations are caused by increased trading volume. Policymakers could enhance futures trading liquidity or restrict speculating positions.

This study examines whether global gold futures returns volatilities and trading activities are threshold cointegrated by using a nonlinear TVECM. The authors detect that some global gold futures returns volatilities and trading activities are threshold cointegrated but some are not. Hence, the findings determine whether the volatility–volume threshold relation holds across countries and investigate the determinants of cross-country differences in different traders.

The accuracy of sensor location estimation influences directly the quality and reliability of services provided by a wireless sensor network (WSN). However, current localization methods may require additional hardware, like global positioning system (GPS), or suffer from inaccuracy like detecting radio signals. It is not proper to add extra hardware in tiny sensors, so the aim is to improve the accuracy of localization algorithms.

The original signal propagation‐based localization algorithm adopts a static attenuation factor model and cannot adjust its modeling parameters in accordance with the local environment. In this paper an adaptive localization algorithm for WSNs that can dynamically adjust ranging function to calculate the distance between two sensors is presented. By adjusting the ranging function dynamically, the location of a sensor node can be estimated more accurately.

The NCTUNs simulator is used to verify the accuracy and analyze the performance of the algorithm. Simulation results show that the algorithm can indeed achieve more accurate localization using just a small number of reference nodes in a WSN.

There is a need to have accurate location information of reference nodes.

This is an effective low‐cost solution for the localization of sensor nodes.

An adaptive localization algorithm that can dynamically adjust ranging function to calculate the distance between two sensors for sensor network deployment and providing location services is described.

The purpose of this paper is to present a novel photomask auto‐correction method for the area‐forming rapid prototyping (RP) system.

A digital light processing (DLP) projector was used in this research as a light source to generate the photomask image. A set of optical lenses were mounted in front of the DLP to rescale the photomask image. The rescaled photomask image was collected into a computer via a camera. By using the technique of image processing, the actual size of the photomask was then calculated. The designed size of the photomask image was eventually achieved by adjusting the relative locations of the lenses.

It was found that this proposed photomask auto‐correction method can produce a more accurate dimension of the photomask image and perform with higher efficiency than the manual calibration processes.

The paper is believed to be the first work to use the image‐processing technique to calibrate the photomask of an area‐forming RP system, as well as to employ a method of adjusting the relative position between the lenses to rescale the photomask image size.

The purpose of this paper is to apply social network analysis (SNA) to study faculty research productivity, to identify key leaders, to study publication keywords and research areas and to visualize international collaboration patterns and analyze collaboration research fields from all Management Information System (MIS) departments in Taiwan from 1982 to 2015.

The authors first retrieved results encompassing about 1,766 MIS professors and their publication records between 1982 and 2015 from the Ministry of Science and Technology of Taiwan (MOST) website. Next, the authors merged these publication records with the records obtained from the Web of Science, Google Scholar, IEEE Xplore, ScienceDirect, Airiti Library and Springer Link databases. The authors further applied six network centrality equations, leadership index, exponential weighted moving average (EWMA), contribution value and k-means clustering algorithms to analyze the collaboration patterns, research productivity and publication patterns. Finally, the authors applied D3.js to visualize the faculty members' international collaborations from all MIS departments in Taiwan.

The authors have first identified important scholars or leaders in the network. The authors also see that most MIS scholars in Taiwan tend to publish their papers in the journals such as Decision Support Systems and Information and Management. The authors have further figured out the significant scholars who have actively collaborated with academics in other countries. Furthermore, the authors have recognized the universities that have frequent collaboration with other international universities. The United States, China, Canada and the United Kingdom are the countries that have the highest numbers of collaborations with Taiwanese academics. Lastly, the keywords model, system and algorithm were the most common terms used in recent years.

This study applied SNA to visualize international research collaboration patterns and has revealed some salient characteristics of international cooperation trends and patterns, leadership networks and influences and research productivity for faculty in Information Management departments in Taiwan from 1982 to 2015. In addition, the authors have discovered the most common keywords used in recent years.