在台灣證券市場中，有許多的技術分析方法或指標，市場參與者或財 務學者會利用歷史資料來做回溯測試，找出可運用的方法或指標，以此來 推測出台股加權指數未來的趨勢，也有學者利用類神經網路(Artificial Neural Network, ANN)考慮經濟景氣、技術分析指標等作為輸入變數來預測 台股加權指數，而本文則利用 EEMD(Ensemble Empirical Mode Decomposition)拆解出來的結果作為 ANN 的輸入變數，並將 ANN 預測出 的值轉換成 FK (Forward-calculated %K) 值，再搭配不同的交易方式，來 補捉台股加權指數的走勢，並比較各種交易方式的績效，找出一個能夠穩 定獲利的交易模型。

Influenza is a public health problem that causes serious illness and deaths all over the world. Vaccination has been shown to be the most effective mean to prevent infection. The primary component of influenza vaccine is the weakened strains. Vaccination triggers the immune system to develop antibodies against those strains whose viral surface glycoprotein hemagglutinin (HA) is similar to that of vaccine strains. However, influenza vaccine must be updated annually since the antigenic structure of HA is constantly mutation. Hemagglutination inhibition (HI) assay is a laboratory procedure frequently applied to evaluate the antigenic relationships of the influenza viruses. It enables the World Health Organization (WHO) to recommend appropriate updates on strains that will most likely be protective against the circulating influenza strains. However, HI assay is labour intensive and time-consuming since it requires several controls for standardization. We use two machine-learning methods, i.e. Artificial Neural Network (ANN) and Logistic Regression, and a Mixed-Integer Optimization Model to predict antigenic variety. The ANN generalizes the input data to patterns inherent in the data, and then uses these patterns to make predictions. The logistic regression model identifies and selects the amino acid positions, which contribute most significantly to antigenic difference. The output of the logistic regression model will be used to predict the antigenic variants based on the predicted probability. The Mixed-Integer Optimization Model is formulated to find hyperplanes that enable binary classification. The performances of our models are evaluated by cross validation.

No description available.

Guided ultrasonic detection system has the ability to inspect long range and not accessible pipelines. Especially, the T(0,1) mode guided wave was used widely at the detection, because the property of non-dispersive. For rapidly judge common features on pipe, this thesis makes an artificial neural network diagnosis system to separate and recognize the signals on pipeline. In the experimental setup, the torsional mode signal are excited by using an array of transducers distributed around the circumference of the 6-inch standard pipe, and the reflected signals contain flange, weld, elbow, and defect on elbow. These features are extracted and have been further processed to limit the size of the neural network; then, the feature signal classify as axisymmetric called black, non-axisymmetric called red, and dividing between the two called R/B ratio. The research also uses finite element method to simulate the weld by building up different kind of profile to analyze its amplitude and simulate the flange, elbow, and defect on the elbow. Because the reflection waves of simulation are too idealize to be the network data, the training data and validation data are collected from the experimental wave. In the recognition of artificial neural network, the signals were getting from two pipes of industry. One has bitumen on it, which makes signals attenuation. The other has a clear elbow and a notch on elbow. The two-class recognition method successfully separates flange and weld in low frequency; but in high frequency, the weld signal amplitude is close to flange signal, because the signals decay when guided waves pass to bitumen, and this makes the judge become error. Furthermore, the network recognizes defects on elbow, where the signals have 3 peaks and 2 peaks when the elbow has defect on it. The training result shows that the 3 peaks have better convergent than the 2 peaks in the network. Finally, the developed method can recognize those defects on the elbow when the reflection signals have 2 peaks, and when reflection signals have 3 peaks, it could not make a good judge because the network limit by sample data.

In response to organic contaminations pollutating water sources of drinking water, domestic water treatment plants (WTP) were transforming from traditional chlorination disinfecton method to advanced ozone-based disinfection processes. However, the effectiveness of water purification procedures n removing AOC (Assimilable Organic Carbon) and DBPsFP (Disinfection By-Products Formation Potential) can be improved. Additionally, the quality of clean water purified at WTP may deteriorate in the water distribution network for various reasons, primarily resulting from the regrowth of microorganisms in the water distribution pipelines. This study investigates and researches the essential water quality items of effluent before and after the advanced water purification treatment plants and water movement to end users through water distribution networks. The investigation proceeded in four directions: (1) the efficiency of removing AOC from raw water using powdered and granular activated carbon biological systems, and the development of an AOC prediction model based on water quality monitoring items using the AutoNet (6.03) method of the artificial neural network system; (2) removal of the byproducts of disinfection from raw water using powdered activated carbon biological systems; (3) examining the relationship between ozone-based and chlorination-based water disinfection methods by comparing the number of coliform bacteria and total bacteria population in traditional and advanced processing units; (4) regarding the water distribution storage facilities for users, water reservoir towers were examined for water quality sampling and analysis and water tower cleaning frequencies. Regression analysis was performed using SPSS ¡]Statistical Product and Service Solutions¡^ statistical software, with the correlation coefficient denoting the closeness of relationships. We anticipate understanding the water quality situation for current users of tap water, and demands for cleaning frequencies, thereby achieving the purpose of improving drinking water safety. Regarding the efficiency of removing AOC from raw water, the results showed powdered and granular activated carbon biological systems performed well, with the AOC removal rate reaching 53% and 54%, respectively, and the SUVA (Specific Ultraviolet Absorbance) value (showed by UV254/DOC) being reduced by 15-18% and 22-23%, respectively. The correlation analysis of the AOC prediction model shows that the GAC (Granular Activated Carbon) had high predictive and actual value R values (R2 = 0.772) after model regressing, and the PAC (Powder Activated Carbon) had higher predictive and actual value R values (R2 = 0.856) after model regressing as well. That the PAC system AOC prediction model has a slightly higher correlation that may be attributed to water contaminations resulting from domestic sewage, agricultural fertilizers, and livestock excretions. In the use of powdered activated carbon biological systems to remove disinfection byproducts, THMsFP (Trihalomethanes Formation Potential) and HAAsFP (Haloacetic acids Formation Potential) functioned with a certain removal efficiency, with the average effluent concentrations being under the regulatory standard of 80£gg/L, respectively, which reduces carcinogenic risks. Correlation analyses conducted using SUVA on the three water quality concentrations (HAA5FP, HAA9FP, and THMsFP) obtained R2 values of 0.805, 0.820, and 0.823, respectively, indicating high levels of correlation. For the results of microbial assessment using ozone and chlorine to process drinking water, the advanced and conventional WTP achieved a removal rate greater than 99% for microbial removal (coliform bacteria and total bacteria population). The correlation analysis between cleaning frequencies and water quality parameters showed the frequency at which the water reservoirs and towers were cleaned has a significant impact on tap water quality in residential compounds and schools that accommodated more than 100 households or less than 99 households. Higher cleaning frequency (more than four cleanings a year) results in better the water quality.

Power quality has attracted considerable attentions from utilities and customers due to the popular uses of the sensitive electronic equipment. Harmonics, voltage swell, voltage sag, and power interruption could downgrade the service quality. Harmonic currents injected by non-linear loads throughout the network could degrade the quality of services to sensitive high-tech customers such as the science park of Xin-Zhu and Tai-Nan in Taiwan. In recent years, massive rapid transit system (MRT) and high speed railway (HSR) have been rapidly developed, with the applications of wide-spread semi-conductor technologies in the auto-traction system. Swell and sag could occur from thundering, capacitor switching, motor starting, nearby circuit faults, or artificial calamity, and could also attribute to the power interruption. To ensure the power quality, harmonic and voltage disturbances detection becomes important. Fourier transformation is used to analyze distorted waves in the frequency domain, with low-pass filter used to eliminate the fundamental component, and then characteristic harmonics can be detected. The complicated process is difficult to operate in real-time. The method-based processing model with physical harmonic data is needed to simplify the processing architecture. The thesis proposes to use wavelet transformation (WT) and probabilistic neural network (PNN) for power quality and characteristic harmonics detection. Wavelet-probabilistic network (WPN) is first used to extract distorted waves. PNN based processing model will then analyze the harmonic components. Computer simulation shows a simplified model to shorten the processing time in this study.

The paper proposes a model of detection for voltages and harmonics using wavelet-probabilistic network (WPN). WPN is a two-layer structure, containing the wavelet layer and probabilistic network. It uses the wavelet transformation (WT) and probabilistic neural network (PNN) to analyze distorted waves and classify tasks. In this thesis, the field programmable gate array (FPGA) is employed for the hardware realization of WPN. In the implementation process, by the use of the hardware description language, the WPN algorithm has been embedded into the FPGA chip. Firstly, we divide the mathematical formula of basic WPN algorithm into several parts in order to set up each module individually, then we integrate all modules to complete the design of basic WPN algorithm with digital circuits by the bottom-up process.

Power system protection is important for service reliability and quality assurance. Various faults may occur due to natural and artificial calamity. Dispatchers are use the changed statuses of protection devices from the Supervisory Control and Data Acquisition (SCADA) system to identify the fault. To reduce the outage duration and promptly restore power services, fault section detection has to be done effectively and accurately with fault alarms. In this thesis, artificial neural networks (ANN) and Grey Relational Analysis (GRA) are used to develop the restoration schemes for emergency operation in a power system including fault section detection (FSD), restoration strategy(RS), and voltage correction(VC). The optimal power flow (OPF) is responsible for verifying the proposed schemes by off-line analysis. With a IEEE 30-Bus power system, computer simulations were conducted to show the effectiveness of the proposed restoration schemes.

The enterprise finance crisis forecast could provide alarm to managers and investors of the enterprise, many scholars advised different alarm models to explain and predict the enterprise is facing finance crisis or not. These models can be classified into three categories by analysis method, the first is single-variate model, it¡¦s easy to implement. The second is multi-variate model which need to fit some statistical assumption, and the third is Artificial Neural Network model which doesn¡¦t need to fit any statistical assumption. However, these models do not consider the industrial effect, different industry could have different finance crisis pattern. This study uses the advantage of Artificial Neural Network to build the process of the enterprise finance crisis forecast model, because it doesn¡¦t need to fit any statistical assumption. Finally, the study use reality finance data to prove the process, and compare with the other models. The result shows the model issued by this study is suitable in Taiwan Electronic Industry, but the performance in Taiwan architecture industry is not better than other models.

Induction motor is the most popular machine in the industry. It is used extensively in mechanical plants, and it is un- avoidable to have the motor¡¦s electrical and mechanical faults due to continuously operating throughout the year. Faults of motors do not only cause the production line to shut down but also imperil the personnel security. A suitable motor maintenance schedule will be a needed to decrease the machine down time. However, major investment might take up to 90% for equipment, and it would be helpful to have a practicable low-cost supervisory scheme on maintenance. If the faults of machine can be detected correctly and effectively, the maintenance efficiency and dependability could be increased greatly. In the past, researches on fault recognition for Induction motors only concentrated on Spectrum analysis with amplitudes based on a constant load. However, the frequency and amplitude of the spectrum analyzed under different fault conditions are also affected significantly by load variations. Using spectrum amplitudes to recognize motor faults is not sufficient in a practical system. Various types of faults and load conditions will influence the spectrum structure. In order to recognize faults under various load conditions, we must consider band shift and amplitude variation as two major factors. In this paper, we use the methods of frequency axis adjustment, load interval and feature exaction to solve the band shift and amplitude variation problems respectively. After the above-mentioned procedures, efficient features are obtained. We use the Back Propagation Neural Network (BPNN) and General Regression Neural Network (GRNN) to train and recognize fault conditions.