Research Topics:

Machine Learning and Neural Networks:

• Proficient in various machine learning algorithms using Python, familiar with TensorFlow and PyTorch;
• Integrating deep learning into finance to establish neural network models for risk identification, financial distress prediction, fraud detection, and stock price forecasting;
• Unsupervised learning.

Large Language Models:

• Building databases, invoking and fine-tuning large language models for downstream tasks.

Network Analysis:

• Using Python to construct shareholder relationship network models of Chinese listed companies for empirical research on corporate governance.

Environmental Accounting and Auditing:

• ESG;
• Developing neural network models with attention mechanisms for carbon price forecasting; Analyzing the spatiotemporal distribution of carbon;
• Assessing the impact of extreme weather on listed companies.

Quantification research on policy texts Analysis:

• Conduct quantitative analysis of policy documents to study the effectiveness of the policies.

Research

Research on the Impact of Corporate Digital Transformation on Carbon Emissions

Accepted by the Journal of Wuhan Business University(ISSN2095-7955)

• Empirically explored how digital transformation reduces carbon emissions by improving production efficiency and resource allocation.
• Analyzed the moderating effects of environmental management reforms and internal corporate governance.

Research on a Financial Fraud Identification Model by Fusing a Convolutional Neural Network

Revised by PLOS ONE(IF:2.9)

• Proposed a financial fraud identification model that combines Convolutional Neural Networks (CNN) and Support Vector Machines (SVM).
• Analyzed the feature contributions and correlations, and visualized them.
• Compared the performance of CNN-SVM with LR, RF, XGBoost, and SVM, and performed statistical testing.

Shareholder Relationship Networks, Information Advantages, and Corporate Digital Transformation

Revised by Humanities and Social Sciences Communications(IF:3.7)

• Empirically studied the impact of shareholder relationship networks on the digital transformation of Chinese listed companies.
• Found that these networks positively influenced digital transformation through "information advantages," especially among stable institutional investors.
• Weakened the effect in state-owned enterprises due to policy burdens, but compensation incentives strengthened it.

Evaluation of Effectiveness and Impact of Green Agricultural Policies — A Quantitative Analysis Based on China's Policy Texts (2017–2022)

Being submitted to Scientific Reports(IF:3.8)

• Used the SBM-GML model to calculate the green total factor productivity of agriculture in China.
• Established a multiple regression model to conduct a quantitative analysis of China's green agricultural policy from the perspective of policy texts.
• Found that supply-side policy tools have a significant positive impact on agricultural green total factor productivity, while environmental policy tools have the opposite effect.

Assessment of the Effectiveness and Impact of Photovoltaic Industry Policies: A Quantitative Analysis Based on Policy Texts from China (2005–2023)

Being submitted to Sustainability(IF:3.3)

• Quantified and analyzed policies related to China's photovoltaic industry.
• Used the SBM-GML model to measure Green Total Factor Productivity (GTFP).
• Conducted empirical analysis on the impact of policy tools on GTFP.
• Found that both supply-side and demand-side policy tools had a significant positive effect on Green Total Factor Productivity.

Construction of a Disease Prediction Model Based on the XGBoost Model

Course paper

• Analyzed and processed medical data, and constructed a disease prediction model by leveraging the advantages of XGBoost.
• Used SHAP to visualize feature contributions.

The Application of Media Coverage in Financial Fraud Warning—Based on Bayesian Optimized LightGBM Model Construction and Empirical Research

Ongoing

• Analyzed media reports on Chinese A-share listed companies using text mining and NLP techniques.
• Optimized LightGBM parameters with Bayesian methods to construct a financial fraud warning model.
• Found that Bayesian optimization significantly reduced the model's error rate.
• Identified negative media coverage as significantly related to financial distress warnings, considering biases and sentiment.

The Carbon Emission Trading Price Prediction Model for China Based on MAIN-transform-KAN

Ongoing

• Used the MAIN-transform-KAN model for carbon price forecasting.