Huei-Wen Teng

Driven by recent AI regulations requiring transparency in high-stakes decision processes, this study enhances the interpretability and accuracy of credit scor- ing models. We emphasize the inherent transparency of the Logistic Regres- sion model and propose a systematic two-step approach for its improvement. First, we augment the model by incorporating polynomial and interaction features, which boosts its performance to match or exceed XGBoost. Second, a useful yet simple Shapley value-based feature selection method addresses multicollinearity and curates a compact, interpretable feature set, ensuring robustness. Extensively validated across two open-source datasets, our enhanced Logistic Regression model outperforms XGBoost, which plateaus under similar feature enhancements. Furthermore, we reveal previously overlooked critical interaction features, offering new insights into their impact on credit scoring. This approach offers a solution for creating precise, interpretable credit scoring models that meet compliance demands.

This study proposes a hybrid deep learning framework that integrates Long Short-Term Memory (LSTM) networks with Association Rule (AR) mining to improve stock market prediction within the S&P 500. Twenty-two representative firms from eleven GICS sectors are selected, covering the period 2010-2024 with daily OHLCV data enriched by technical indicators such as moving averages, Bollinger Bands, MACD, and volatility measures. The LSTM component captures nonlinear temporal dependencies in stock returns, while the AR layer introduces interpretable cross-market validation signals derived from global indices. Unlike prior studies that depend on high-frequency data or institutional-scale computing, this research highlights the feasibility of implementing hybrid deep learning using personal computers. Empirical evidence demonstrates that the LSTM with AR framework consistently outperforms traditional baselines, including Logistic Regression and Buy-and-Hold strategies, providing enhanced robustness, interpretability, and practical relevance for individual investors in algorithmic trading applications.

Digital assets (DAs) such as cryptocurrencies, tokenized securities, stablecoins, non-fungible tokens (NFTs), and central bank digital currencies, are transforming financial markets with new business models, investment opportunities, and transaction efficiencies. Underpinned by blockchain, distributed ledger technology, and smart contracts, digital innovations are reshaping the financial ecosystem. However, their rapid growth introduces substantial risks, including fraud, market manipulation, cybersecurity threats, and regulatory uncertainty. This position paper offers an interdisciplinary and empirically grounded analysis of the DA landscape. We define and classify major asset types, trace their evolution from speculative instruments to functional tools, and assess current adoption trends. Additional technological developments (e.g., decentralized finance and NFT expansion) are examined for their role in accelerating this transformation. We also analyze the global regulatory landscape, highlighting jurisdictional differences, classification challenges, and emerging governance frameworks. To address key risks, we derive mitigation strategies via quantitative analysis and case-based evidence. The risks include balancing innovation with investor protection through adaptive regulatory design, promoting cross-border regulatory harmonization to prevent arbitrage and fragmentation, and supporting experimentation through regulatory sandboxes and innovation hubs. By adopting a forward-looking, evidence-based, and collaborative regulatory approaches, stakeholders can harness the benefits of DAs while managing systemic risks and maintaining market integrity.

The rapid emergence of digital assets, underpinned by technological advancements such as blockchain, distributed ledger technology (DLT), and smart contracts, has triggered a paradigm shift in the global financial ecosystem. These digital assets, which encompass cryptocurrencies, tokenized securities, stablecoins, non-fungible tokens (NFTs), and central bank digital currencies (CBDCs), hold the potential to transform financial markets by enabling new business models, investment opportunities, and efficient transaction mechanisms. However, their accelerated growth also introduces a unique set of challenges and risks, such as fraud, market manipulation, cybersecurity threats, and regulatory uncertainties. This position paper presents an interdisciplinary, empirical analysis of the digital asset landscape, focusing on the definition and classification of digital assets, their evolution from novelty to necessity, and the current state of adoption and regulation. We explore the various types of digital assets, their unique characteristics and use cases, and the technological innovations that have shaped their development, such as the advent of blockchain technology and the rise of decentralized finance (DeFi) and NFTs. Moreover, we examine the regulatory landscape surrounding digital assets, highlighting jurisdictional approaches, regulatory classifications, and key developments in the space, as well as the challenges and opportunities that regulators face in devising effective regulatory frameworks. To address the risks associated with the proliferation of digital assets, we outline several mitigation strategies and recommendations for regulators, market participants, and stakeholders based on quantitative analysis and empirical findings. These include balancing innovation and risk, by formulating regulations that safeguard the interests of consumers and investors while fostering an environment conducive to innovation; promoting global regulatory coordination and harmonization, to reduce the potential for regulatory arbitrage and enhance cross-border cooperation; and leveraging regulatory sandboxes and innovation hubs, to support the growth of digital asset businesses and facilitate continuous learning and adaptation. By adopting a forward-looking and flexible approach to regulation and engaging in ongoing dialogue with market participants and stakeholders, regulators can ensure that the benefits of digital assets are realized while mitigating the associated risks.

Developing an accurate credit scoring model is crucial in the rapidly growing consumer finance industry. Explanatory variables can be divided into dynamic time-related and static time-unrelated features. In this study, we leverage convolutional neural network (CNN) and recurrent neural network (RNN) to extract new features from dynamic data and combine them with static variables. Our results reveal two key insights. Firstly, the addition of CNN- or RNN-extracted features to four machine learning algorithms, namely Logistic Regression, Decision Tree, LightGBM, and Neural Networks, leads to marginal improvements in AUC and F1-Score performance. Secondly, while LightGBM performs better than other algorithms in the training dataset, logistic regression performs comparably or better in the testing dataset.

Credit scoring models predict a person's creditworthiness. Accurate and interpretable credit scoring models help maximize a financial institution's risk-adjusted return and hence are of considerable importance. Although the XGBoost is the state-of-art classifier for this task, its complexity makes it challenging to interpret. This paper proposes a novel approach, the rescaled cluster-then-predict method, which rescales features before clustering the data into subgroups. Our approach leads to a more interpretable, yet still competitive, Logistic Regression that even outperforms XGBoost in some cases. Empirical results on two datasets highlight the benefits of the proposed method. We also find that clustering positive cases (rather than all cases) leads to similar results while requiring substantially less computation time. These findings have important implications for the finance industry, where accurate, interpretable, and efficient credit scoring models are in high demand.