Chen C.-Y.,National Pingtung University |
Shyue S.-W.,National Sun Yat - sen University |
Chang C.-J.,Transworld Institute of Technology
International Journal of Innovative Computing, Information and Control | Year: 2010
Data mining is commonly used in a wide range of different fields. It is an important tool for analyzing information in research areas where there is a huge collection of data. The aim of this study is to determine the various patterns that characterize multi- dimensional marine environments. In particular, we wish to determine the association patterns, the characterize Taiwan's Dapeng Bay. The processes of association rule (AR) mining and decision tree (DT) analysis are the main methodologies used in this study. We also utilize Weka, a comprehensive suite of Java class libraries for implementing many machine learning algorithms and the Clementine 10.1 software package for data mining analysis. Applications constructed from the AR model and decision tree model can be applied in other scientific domains. This paper describes the data mining process for Dapeng Bay, which is located in the south of Taiwan. ICIC International © 2010 ISSN.
Cheng J.-H.,National Yunlin University of Science and Technology |
Chen H.-P.,National Yunlin University of Science and Technology |
Lin Y.-M.,Transworld Institute of Technology
Expert Systems with Applications | Year: 2010
One of the major difficulties in investment strategy is to integrate supply chain with finance for controlling the marketing timing. The present study uses not only the different indexes in fundamental and technical analysis, but also the rough set theory and artificial neural networks inference system to construct three investment market timing classification models. This includes probabilistic neural network classification model, rough set classification model and hybrid classification model combining probabilistic neural network, rough sets and C4.5 decision tree. We use the forecasting accuracy and investment return to evaluate the efficacy of these three classification models. Empirical experimentation shown hybrid classification model help construct a better predictive power trading system in terms of stock market timing analysis. © 2009 Elsevier Ltd. All rights reserved.
Lin J.-C.,Transworld Institute of Technology
JP Journal of Heat and Mass Transfer | Year: 2010
In this study, performance analysis and ecological optimization for an Atkinson engine with constant-temperature reservoirs are conducted. The ecological objective function is defined as the power-output minus the loss rate of availability. The parameters that maximize the ecological objective function and make the performance optimum are investigated. The results show that a maximum ecological function (MEF) exists and the corresponding power-output at the MEF condition is slightly lower than the maximum power-output. With the increase of the number of transfer units of the hot-side heat exchanger NH (or the cold-side heat exchanger NL), the MEF and the power-output at the MEF condition increase. The thermal efficiency at the MEF condition is lower than the maximum thermal efficiency that can be obtained. For a fixed NH, the thermal efficiency at the MEF condition increases dramatically at lower NL, while it increases slightly at larger NL. In addition, the thermal efficiency at the MEF condition gets larger as NH becomes greater. When the sum of NH and NL is fixed, the variation of the thermal efficiency at the MEF condition for various NH/(NH + NL) ratio is not obvious.
Liu M.-S.,Academia Sinica, Taiwan |
Li H.-C.,Academia Sinica, Taiwan |
Chang Y.-M.,Transworld Institute of Technology |
Wu M.-T.,Taiwan Agricultural Research Institute |
Chen L.F.O.,Academia Sinica, Taiwan
Plant Science | Year: 2011
Our previous study revealed a cytokinin-related retardation of post-harvest floret yellowing in transgenic broccoli (Brassica oleracea var. italica) that harbored the bacterial isopentenyltransferase (ipt) gene. We aimed to investigate the underlining mechanism of this delayed post-harvest senescence. We used 2D electrophoresis and liquid chromatography-electrospray ionization-mass spectrometry/mass spectrometry for a proteomics analysis of heads of ipt-transgenic and non-transgenic inbred lines of broccoli at harvest and after four days post-harvest storage. At harvest, we found an accumulation of stress-responsive proteins involved in maintenance of protein folding (putative protein disulfide isomerase, peptidyl-prolyl cis-trans isomerase and chaperonins), scavenging of reactive oxygen species (Mn superoxide dismutase), and stress protection [myrosinase-binding protein, jasmonate inducible protein, dynamin-like protein, NADH dehydrogenase (ubiquinone) Fe-S protein 1 and stress-inducible tetratricopeptide repeat-containing protein]. After four days' post-harvest storage of non-transgenic broccoli florets, the levels of proteins involved in protein folding and carbon fixation were decreased, which indicates cellular degradation and a change in metabolism toward senescence. In addition, staining for antioxidant enzyme activity of non-transgenic plants after post-harvest storage revealed a marked decrease in activity of Fe-superoxide dismutase and ascorbate peroxidase. Thus, the accumulation of stress-responsive proteins and antioxidant enzyme activity in ipt-transgenic broccoli are most likely associated with retardation of post-harvest senescence. © 2011 Elsevier Ireland Ltd.
Yang S.-F.,Transworld Institute of Technology
International Journal of Robust and Nonlinear Control | Year: 2011
This paper presents an efficient algorithm for the generation of QFT bounds for robust tracking specifications for plants with affinely dependent uncertainties. For a plant with m affinely dependent uncertainties, it is shown that whether a point in the Nichols chart lies in the QFT bound for a robust tracking specification at a given frequency can be easily tested by computing the maxima and minima of m2m-1 univariate functions corresponding to the edges of the parameter domain box. This test procedure is then utilized along with a pivoting procedure to trace out the boundary of the QFT bound with a prescribed accuracy or resolution. The developed algorithm has the advantages that (1) it is efficient in the sense that it requires less floating point operations than other existing algorithms in the literature; (2) it can avoid the unfavorable trade-off between the computational burden and the accuracy of the computed QFT bounds that has arisen in the application of many existing QFT-bound generation algorithms; (3) the maximum allowable error of the computed QFT bound can be prespecified; and (4) it can compute QFT bounds with multi-valued boundaries. Numerical examples are given to illustrate the proposed algorithm and its computational superiority. © 2010 John Wiley & Sons, Ltd.