34254 research outputs found
Sort by
Application of Machine Learning in Credit Card Fraud Detection: A Case Study of F Bank
No full textDue to the Covid-19 pandemic, people are increasingly engaging in non-face-to-face credit card transactions in their daily lives. However, this trend has also provided opportunities for malicious actors to obtain customer credit card information through various illicit means, leading to a continuous rise in credit card fraud. Traditional fraud detection methods, relying on extensive rules and manual judgment, struggle to effectively prevent the evolving techniques of fraud and often result in significant false positives, requiring substantial time for transaction verification.
In recent years, the development of big data and machine learning algorithms has offered an effective solution to this challenge. This study employs three common machine learning algorithms\ue2Logistic Regression, Random Forest, and Extreme Gradient Boosting\ue2for predicting credit card fraud. Utilizing transaction data from Bank F time period from January 2021 to May 2023, including fields such as transaction ID, credit limit, occupation, transaction date, transaction time, transaction amount, etc., the study addresses the issue of imbalanced data in credit card fraud through sampling methods. Different ratios of normal to fraud samples, coupled with varying sampling frequencies, are employed along with ensemble learning techniques to enhance the accuracy and stability of the predictive model. Subsequently, various commonly used machine learning evaluation metrics are applied to identify the best model.
The empirical results indicate that the Extreme Gradient Boosting model performs best in detecting credit card fraud. In scenarios with different sampling ratios of normal to fraud samples, the study identifies key features such as changes in the cardholder's transaction behavior concerning transaction region, frequency, and amount. The results of this study provide the bank with references on how to develop more effective strategies for fraud prevention
A Study on the Conflict of Interregional Jurisdiction between Hong Kong and Mainland China\uef\ubcWith the Focus on Judicial Practice of Civil and Commercial Cases
No full textOn July 1, 1997, the return of Hong Kong to Chinese sovereignty took place and there has been \ue2 One country, Two systems\ue2 between China and Hong Kong. The two regions are under different legal systems, among which Hong Kong retained its previous common law system and China uses socialist legal system. Further observation of the development between China and Hong Kong has left the question whether Hong Kong can continue to keep its independence in legal processes based on \ue2The Basic Law\ue2 without clear answers.
\ue3For instance, a party may face lawsuits of concurrent proceedings in Hong Kong and in China and most such examples are civil and commercial cases. In other words, the differences that exist between Hong Kong and China regarding their theories and design of laws would influence courts of both regions, when they are met with parallel proceedings, including their mentality and basic principles used in adjustment. Since parallel proceedings due to the conflict of laws are inevitable, how the courts from both regions deal with their jurisdiction under competition is crucial. Therefore, upon the conflict of laws, whether and how courts from both sides apply the doctrine of forum non conveniens, consent jurisdiction, anti-suit injunctions, etc. to regulate the adjustment mechanism, narrowing their gap in actual practice and results, or the other way around, would greatly affect the rights of all parties.
\ue3Hence, made its studies based not only on lagal district concept which define the conflicts between the two regions as Conflict of Interregional Jurisdiction, but also use the theory bases of International Jurisdiction and Interregional Jurisdiction as the startings point of research to analize the comprehensive collections of typical cases in Hong Kong and China that involve conflict of interregional jurisdiction in area of civil and commercial cases. The paper analyses how the courts from different regions confirm their jurisdiction and how they apply the aforementioned conflict of interregional jurisdiction so as to find out their mentality and the ground where they stand, when it comes to conflict in civil and commercial cases.
\ue3After examing the selected cases, it is clear that conflict of jurisdiction has been natural, yet in the aspect of actual practice, courts from both regions basically showed the mindset of consent jurisdiction or even expansion of their own jurisdiction. In terms of their practice of the above regulating adjustment mechanism, first of all, the Hong Kong court is based on three stages test to verify which court is the forum non conveniens. As to the Chinese court, it rarely uses Doctrine of Forum Non Conveniens. Furthermore, courts from both sides fully respect the autonomy of parties and the effectiveness of consent jurisdictions; both treat its nature in principle as exclusive jurisdiction, prioritizing it over the adjustment mechanisms. Lastly, the Hong Kong court built up reviewing requirements of anti-suit injunctions based on its past cases, and held the principle of international comity, requiring extremely high standard for inspection, avoiding infringing on foreign jurisdiction. In comparison, the Chinese court for a long time has turned a blind eye on the anti-suit injunctions. However, since issuing the maritime injunction of the first maritime case as well as in recent years, the cases involved with intellectual property rights, there have been records of issuing anti-suit injunctions or anti-anti-suit injunctions, consenting its protection of Chinese intellectual property rights; therefore becoming more relaxing on issuance of such anti-suit injunctions.
\ue3In summary, this paper assumed at first that the conflict of concurrent jurisdiction on civil and commercial cases was a result of Hong Kong\ue2s returning to China in 1997 and the existence of separate legal districts of judicial. Through practical examinations and analysis of multiple cases, and review of practice of courts from both regions, the above results have been achieved
A High-Linearity and High-Resolution Wide Range Temperature Detector and A High-Performance Low-Power Neural Network Hardware Architecture
No full textAs neural networks flourish, many sophisticated models are emerging. Traditional hardware architectures struggle to achieve high throughput and low power at the same time. Consequently, hardware architectures meticulously designed for neural networks have become an indispensable key to resolve the mentioned issues. Ensuring stability and reliability during any operation, the on-chip sensors play a pivotal role. The first topic in this thesis presents a novel architecture for temperature sensing, enhancing the detectable temperature range while achieving high linearity and high resolution. Additionally, thanks to the support of TSMC JDP project, this thesis secondly presents an efficient hardware accelerator optimized for convolution neural networks (CNN) and an innovative photonic
neural network architecture.
The high-linearity and high-resolution wide range temperature detector employing PTAT and CTAT current generators is coupled to a current-to-frequency converter composed of differential delay cells. Two frequency sets with opposite temperature characteristics are attained. By the division of these frequencies, common-mode errors are reduced enhancing the differentiation in temperature variations. The proposed sensor can measure temperatures range from -40 \ue2\ua6C ~ 100 \ue2\ua6C with 0.5 \ue2\ua6C resolution. Maximum linearity error of 0.436 % is found by on-silicon measurement.
The second topic includes two systems: a high-performance neural network hardware accelerator and a photonic neural network architecture. The first system presents a novel hardware architecture specifically designed for convolutional computations. Besides a highly parallelized PE array, the hardware accelerator is featured with a reshape module for memory data processing, which allows synchronous reading and computation. The on-silicon measurement shows the performance of 54.61 GOPS at 96.35 mW. The second system presents an innovative photonic neural network (PNN) structure. Take advantage of the high-speed and low-power consumption advantages of optical devices, extensive convolutional computations combined with feature extraction is realized. Based on 10,000 MNIST benchmark testing, the recognition rate of 93.8% is achieved
Development of Key Photonic Devices for Co-packaged Optics Applications
No full textThis thesis is dedicated to develope key components of silicon photonics for Co-packaged Optics (CPO). We designed silicon photonics components include multiplexers, demultiplexers, and optical modulators for Wavelength Division Multiplexing (WDM).
This thesis employs formulas to design low-loss bend waveguides. Low-loss bends have lower propergation loss than circular bends because of its gradual variation width and radius. In Addition, we use the method of integral to calculate the phase of low-loss bends, since it is not easy to calculate like circular bends. By using this method, we can spend less simulation time than using FDTD software.
In this thesis, we also employ the low-loss ring modulator which is comprised of doping low-loss bends. We construct nine low-loss ring modulators and vary the length of phase shifters and gap spacings for comparison. The resultant of the low-loss ring modulator with 100% circumference phase shifter and 180 nm gap spacing enables a tuning efficiency of 35.25 pm/V, a quality factor of 5800, and an electro-optic bandwidth of 45.33 GHz. Under 50 Gbps NRZ signal, Optical Modulation Amplitude (OMA) and Signal-to-Noise Ratio (SNR) is 28.54 mV and 7.78 dB, respectively.
We explored how the process variation would impact the characteristics of Mach-Zehnder Interferometer (MZI). By using formulas and conducting simulations, we find the relationship between the free spectral range (FSR) and interference wavelength shift. The effects of two arm waveguides and Common waveguides are discussed separately.
We compare three types (de)multiplexers composed of ring resonators and MZIs by simulation software. According to the measurement results, the MZI cascaded ring resonator one has approximately 7 dB loss. We think it can be improved by replacing them with low-loss ring resonators. The crosstalk is approximately -30 dB, it is consistent with simulation results.
Finally, an eight-channel optical transceiver featuring ring resonator modulators and high-speed photodetectors on each channel in the chip is packaged
An LED Driver for Motorcycles
No full textThis thesis proposes a Light-Emitting Diode (LED) driver circuit designed for motorcycle lighting. This circuit architecture features both step-down and step-up functionalities to accommodate input and output variations. By utilizing feedback from the output current signal and employing a digital signal processor to calculate the switch duty cycle, it adjusts the voltage gain of the converter to meet the required output voltage. The thesis employs simulation software SIMPLS to obtain loop gain and the phase of the circuit, analyzes the resulted Bode plot, and designs suitable compensators to achieve system stability. Ultimately, a 30W rated drive circuit for a series connection of 10 LEDs was constructed. Experimental results demonstrate the driver's capability to handle variable inputs and outputs\ue2input variations ranging from 8V to 13.5V from the battery, and output variations from 1 to 10 LEDs (3V~30V). The driver's operational mode switches based on the required output rated voltage, allowing for adjustment of output voltage and current, the maximum efficiency can achieve 98.3%
Tunable Activated Supercapacitor Electrodes Based on Conjugated Microporous Polymers
No full textThe current thesis highlights the synthesis of conjugated microporous polymers and then tuning their performance as supercapacitor electrodes. In the first chapter we introduce a recent literature survey of modified porous organic polymers. Organic porous polymeric materials are distinguished as their particular surface areas are high, modulated porosities, minimal weights, in addition to outstanding thermal and chemical stabilities. These inspiring features are usually derived due to the wide variety of their building units, \ucf-conjugated extension, in addition to their incorporation of organic atoms, nitrogen, as well as carbons. Organic porous polymeric materials are distributed into four important classes as covalent organic framework (COF), covalent triazine framework (CTF), microporous conjugated polymer (CMP), hyper-crosslinked polymer (HCP), as well. Realistically, those CMPs apply widely in intensive fields, for example, gas division, or capture, energy conversion, energy storing, photocatalysis, sensors, aqua desalination, drug delivery, and optical apparatus. However, energy storage gained enormous interest in recent decades thus we also exhibit in this chapter some limitations of CMPs as supercapacitors electrodes. We focus this work on design of novel CMPs and then investigating their validities toward this important application, and then we target such improving, and modulating their electrical storage.
In the second chapter; we execute Monopot interaction and polycondensation between borono ester of triphenylamine (TPA-3Bor) with bromo-benzene (Bz-4Br), and bromo-pyrene (Py-4Br) to design modern couple of TPA-based CMPs respectively of TPA-Bz, and TPA-Py CMPs. These two CMPs show considerably chemical durability, high surface areas, and high thermal stability. TPA-Bz, as well as TPA-Py CMPs represent dynamic and stable performances when employed as supercapacitor electrodes with high capacitance up to 78 F g-1 at current density 1 A g-1 in addition to their high stability even after 2000 cycles. Most importantly, designing such these couple of CMPs not only displayed high effective materials for supercapacitors but also display the importance of conjugation for future designed materials.
In the third chapter, High porosity CMPs with a regular porous structure have emerged as fruitful electrode candidates for supercapacitors, however they don't conduct electricity well so restrict how effectively they can be used and how well they operate as supercapacitor materials. The design of such three-dimensional networked CMPs cored conductive polypyrrole (PPy) nanotubes (PNT) boosts the total electrical conductivity and capacitive storage of the former. The highly conductive PNT grants CMPs an improvement of electron transportation between their particles. In the current study, novel TPA-TPA, and TPA-Th CMPs are designed, and characterized then we hold compare between them and their composites with PNT. PNT nanocomposites show improved surface areas, and thermal stability (char yield) than pristine PNT or CMPs up to 210.78 m2g-1 and 71.1 % respectively. The specific capacitance of pristine the TPA-TPA CMP and TPA-Th CMP electrodes increase from 41.82, and 109.8 F g-1 up to 131, and 292.6 F g-1 respectively in TPA-PNT, and Th-PNT nanocomposites. The significance of CMPs-PNT nanocomposites as supercapacitors electrodes is far above other CMPs derivatives elucidate the importance of in situ composite design between PNT and CMPs on the nanoscale. Ratio modification between CMPs to PNT offers consequent modulation of physical, chemical features, specific capacitance, and capacitance retentions of the nanocomposites hence considers a tradeoff point.
In the fourth chapter; we synthesize and characterize couple conjugated microporous polymers (CMPs), TPA-QP and TPA-TBP CMPs, incorporating various redox active units of phenazine dibenzo[5,6:7,8]quinoxalino[2,3-i]phenazine (QP), and tetrabenzophenazine (TBP) respectively then demonstrate their activities as modern supercapacitor electrodes. The physio and electrochemical features of these CMPs vary depending on the natures of incorporated phenazine units, as revealed experimentally and computationally. TPA-QP, and TPA-TBP CMPs offer tunable capacitance performance depending on number of adsorption sites of electrolyte cations.
In the fifth chapter; We present the synthesis of three redox active CMPs based phenazine of QP coupled with various linkers of Py, 2,4,6-trisphenyl) pyridine (TPP), and 2,4,6-trisphenyl)-1,3,5-triazine (TPT). These CMPs display high thermal stability with char yields of 82.5%, as well as improved surface areas of 527.22 m2g-1. Interestingly, variation of subunits with phenazine core derive tunable energy storage performance as well as modulate storing mechanism. QP CMPs show high specific capacitances as well as energy densities respectively of 121.9 F g-1, and 16.93 Wh Kg-1 respectively. Further, they show high capacitance retention which recorded 96% after 5000 cycles, and their capacitances are comparable to other earlier reported porous materials
The Relationship between in Top Management Team Composition and Organizational Performance: The Mediating Role of Employee Turnover Rate
No full textIn the post-pandemic era, enterprises in the high-tech industry are confronted with numerous new challenges and changes. The impact of the pandemic and market uncertainties compels companies to be more agile and flexibility, ensuring competitiveness and operational efficiency. Enterprises not only need to respond to environmental changes but also need to transition towards establishing a model of business development that prioritizes employee well-being. In this context, this study aims to delve into the intricate relationship between employee turnover, top management team (TMT) heterogeneity, and organizational performance through the lens of Upper Echelons Theory. Amidst the continuous evolution of the business environment, employee turnover is considered a crucial factor influencing the structure of top management teams and organizational performance. By exploring the complex interactions between these variables. The study utilized publicly available data from listed companies in the passive component industry, employing SPSS and Mplus for analyses. The findings revealed a negative relationship between TMT average tenure and employee turnover, a positive relationship between TMT turnover frequency and employee turnover rate, a positive association between employee turnover and organizational performance, and mediating effects of employee turnover rate in the relationships between TMT average tenure, TMT turnover frequency, and organizational performance. In conclusion, this study presents managerial implications based on empirical results and provides valuable insights for future research endeavors
Cross-domain Fine-grained Access Control Framework for IoT data based on Metaverse Environment
No full textIntegrating IoT and metaverse creates a new model that connects the physical and virtual worlds through IoT sensors. These sensors collect real-time environmental data and seamlessly integrate it into the virtual environment. This allows users to monitor and control the physical world in the virtual environment, making it an ideal platform for testing new technologies and scenarios without incurring high costs. However, managing and sharing this data presents challenges, including different public key settings, the dynamic nature of cross-domain data access, and the traversal of digital entities in multiple virtual worlds. These challenges make cross-domain access control more complex. Traditional encryption solutions operate under a single authority and need help accommodating IoT device owners from different domains, leading to insecure cross-domain data access and high costs. The thesis proposed an integrated IoT and metaverse access control framework to address these challenges. It empowers data owners to securely share encrypted data with specific users across domains through fine-grained authorization. It also allows data owners to maintain control over data sharing while ensuring the authenticity and integrity of the data in the encryption system. The proposed authorization protocol includes authorization proof and revocation functions based on zero-knowledge proof. This allows the data owner to prove that the authorization has been granted to the user and to revoke the authorization in case of a dispute, thereby increasing the data owner's control. The proposed framework is a comprehensive solution for achieving secure, fine-grained cross-domain data access control. It provides a more efficient and cost-effective approach for organizations operating at the intersection of IoT and metaverse
The Effect of Subordinate-Supervisor Fit of Promotion Opportunities and Proactivity Behavior for Civil Servants
No full textIn Taiwan, the promotion system for civil servants is protected by law, and they benefit from being a civil servant. One of the benefits is lifetime commitment. They seem to have stable promotion opportunities. However, the promotion opportunities for civil servant are not easy as imagined. Because the relationship with supervisor affects the possibility of promotion opportunities. This study uses social exchange theory as a framework to explore the relationship between \ue2Person Supervisor Fit\ue2,\ue2 Promotion Opportunities\ue2, and \ue2Proactive Behavior\ue2, Furthermore, it investigates the mediator effects between \ue2Person Supervisor Fit\ue2 and \ue2Proactive Behavior\ue2, and the moderating effects of \ue2Leader-Member Social Exchange\ue2, \ue2Person Supervisor Fit\ue2, and \ue2Promotion Opportunities\ue2.
This research employed an online questionnaire distributed to employed civil servant from public apartment in Taiwan. The survey was conducted in three waves, each separated by a two-week interval, resulting in a total of 404 valid responses. The data is analyzed with the SPSS 26.0 hierarchical regression analysis (Raudenbush & Bryk, 2002) to test the research hypotheses. The research results find that promotion opportunities completely mediate the relationship between personal supervisor fit and proactivity behavior, and the person supervisor fit is positively related to promotion opportunities. Also, there is a positive relating between promotion opportunities and proactivity behavior. Finally, the hypotheses of leader member social exchange moderate the relationship between person supervisor fit and promotion opportunities was not supported by the data. Based on the research results, this study proposes practical management implications and future research directions for reference by subsequent researchers
A case study on the recruiting activities and outcomes of a head-hunt rookie
No full textThe existing research on talent acquisition encompasses four main categories: the impact on clients, organizational behavior and work of recruitment personnel, job-seeking behavior of candidates, and the operational aspects of recruitment agencies. This study specifically focuses on novice recruitment professionals, aiming to explore the actual recruitment activities and outcomes in the cases they handle, with the objective of understanding the development of rookie recruiters.
In this research, data were collected from the handling of 24 candidate recruitment cases by a novice recruiter. The study summarizes the operational processes of recruitment activities and communication correspondences with candidates. Based on the actual outcomes of the recruitment process, candidates are categorized into three groups: those not recommended after contact, those recommended but declined interview, and those recommended but deemed unsuitable by the employer. Additionally, candidates who rejected the job offer are also included in the analysis