Menu
Home
Google Scholar
GitHub
CSDN
LinkedIn
Twitter

王子源 Ziyuan Wang

alt text 

Quantitative Analyst, PhD,
Research Department,
Zheshang Fund.
Lujiazui Road No.99 10L, Pudong New District,
Shanghai, China, 200433
English CV | 中文简历
E-mail: wangziyuan AT zsfund.com

About Me

I am currently a Quantitative Analyst 👨‍💻 at Zheshang Fund , got PhD 🎓in SUFE AI Lab in 2024, supervised by Professor Hailiang Huang. My main research interests include Deep Learning, Machine Learning, Natural Language Processing and Quantitative Investment, focus on using AI to solve practical problems. Before SUFE, I received the B.S. degree in Statistics from Hunan University in 2018.

Research🔬

My research interests include:

  • Deep Learning

  • Natural Language Processing

  • Text Classification

  • Semantic Textual Similarity

  • Quantitative Investment

Publications📃

  1. Z. Wang, H. Huang and S. Han, "IDEA: Interactive Double Attentions from Label Embedding for Text Classification", 2022 IEEE 34th International Conference on Tools with Artificial Intelligence (ICTAI), Macao, China, 2022, pp. 233-238, doi: 10.1109/ICTAI56018.2022.00041. CCF-C, CORE-B, Acceptance rate 15.7%.

  2. Guo, B.*, Zhang, X.*, Wang, Z.*, Jiang, M.*, Nie, J.*, Ding, Y., Yue, J., & Wu, Y. "How Close is ChatGPT to Human Experts? Comparison Corpus, Evaluation, and Detection". ArXiv, abs/2301.07597.(Accepted by LLM@IJCAI 2023 recently!)

  3. Z. Wang and Y. Wu, "Investigating the Effectiveness of Whitening Post-processing Methods on Modifying LLMs Representations," 2023 IEEE 35th International Conference on Tools with Artificial Intelligence (ICTAI), Atlanta, GA, USA, 2023, pp. 813-820, doi: 10.1109/ICTAI59109.2023.00124. CCF-C, CORE-B, Acceptance rate 21%.

  4. Wei J, Wang Z, Du Q, et al. New model of utility analysis and performance prediction in crowdfunding: A perspective of behavior-related decision[J]. Expert Systems with Applications, 2024: 124102. doi: 10.1016/j.eswa.2024.124102, JCR Q1 and SCI I zone top journal.

Note: * indicates equal contribution.

Working Experience🧱

  1. Zheshang Fund, Quantitative Researcher, 2024.3 to now

    • Stock selection by Graph neural network based on text information. According to the text content, the correlation between A-share stocks is explored, and the potential market dynamics are revealed through text analysis. This paper studies several SOTA text representation models such as FinBERT, ERNIR3.0 and DMETA, and combines Whitening and other techniques to improve the anisotropic representation of language models.

    • The investment strategy based on GCN, GAT and other graph neural network methods to establish the connection with the stock price, based on the volume and price characteristics, the Out-Of-Sample test IC reaches 14.44%, RankICIR 13.94%, ICIR, RankICIR exceeds 1.2 in the 2021-2023 interval of China Securities Index. The effect is better than the baseline model and the graph information at the same model performance.

  2. Millennium Management, CRTC, Junior Quantitative Analyst, 2023.7 to 2023.12

    • A quantitative investment application based on NLP in Earnings Conference Call sentiment factors. Through text analysis of earnings conference calls of listed companies in the US stock market, based on FinBERT, FLANG-BERT, RoBERTa and FinGPT and other large language models, explore a variety of deep learning training methods such as language model inference, fine-tuning and upstream pre-training + downstream fine-tuning. Moreover, the sentiment analysis and inference of different Prompts in LLMs are compared and analyzed. The Ensemble factor is predicted by using the above models; Based on the research results of SOTA in academia and industry, the NLP feature factors based on behavioral finance are established. Combined with the above factor signals, we built Pipeline framework for market investment sentiment mining.

    • Achieved significant results,outperforming the returns of factors provided by data vendors. Possess extensive knowledge and practical application insights into the use of large language models like ChatGPT in quantitative investing, demonstrating a cutting-edge and innovative approach.

  3. On-Campus Employment

    • Academic Reviewers: Applied Intelligence,International Journal of Computer Sciences and Engineering, ACM International Conference on Information and Knowledge Management

    • Teaching Assistant: Advanced Machine Learning(Fall 2019), Application of Big Data in Economics (Fall 2021)

    • Research Assistant: Fall 2020, Fall 2021, and Spring 2022 semesters.

Projects💻

  1. How Close is ChatGPT to Human Experts? Comparison Corpus, Evaluation, and Detection 2022.12-2023.07

    • The earliest to open source comparison datasets and detector models in academia and industry

    • We collected human-ChatGPT comparative data in an open-domain Q&A task and published the first ChatGPT Corpus, HC3 (Human ChatGPT Comparison Corpus). We conducted Turing tests, textual statistics, sentiment analysis, and linguistic analysis on the corpus. We summarized the textual paradigms of ChatGPT and its differences from human responses, and developed a series of ChatGPT content classification detectors based on single texts and Q&A pairs using deep learning and machine learning methods, achieving significant detection results.

    • Our detector demo has received over 300,000 global visits, with the open-source model averaging 300,000+ monthly downloads, the dataset averaging 30,000+ monthly downloads, and acquiring over Github stars and 300+ paper citations.

    • Accepted by LLM@IJCAI , Co-First Author.

    • Visiting Our Demo: ChatGPT detectors 🔥

  2. Investigating the Effectiveness of Whitening Post-processing Methods on Modifying LLMs Representations, 2022.09-2023.06

    • Investigated the decorrelation effects of vector matrix whitening methods such as PCA and ZCA on the textual representations of large language models. This research aimed to standardize the text representation learning ability of language models, addressing the traditional models' oversight of the assumption of orthogonal bases in cosine similarity. The study significantly improved performance across nearly 20 datasets in different NLP tasks, applying to models including Bert, DistilBert, and GPT-2.

    • Published in ICTAI'2023, First Author.

  3. New Hybrid Model of Utility Analysis and Performance Prediction in Crowdfunding: A Perspective of Behavior-related Decision, 2022.08-2023.02

    • In this work, the prospect theory (PT) is introduced to capture the heterogeneity of investors' support behavior in crowdfunding. The empirical results show that there is a significant correlation between behavioral factors and crowdfunding performance, and the AUC of classification increases by 7.39% on average.

    • Published in ESWA (Expert Systems With Applications), JCR Q1 and SCI I zone top journal. Corresponding Author.

  4. IDEA: Interactive Double Attentions from Label Embedding for Text Classification, 2022.06-2022.09

    • Proposed a succinct method to enhance BERT's performance in text classification by utilizing label embedding techniques in supervised learning. Developed a novel model structure that integrates text and label information on top of the existing BERT model, achieving significant improvements on multiple public datasets.

    • Published in ICTAI'2022 , First Author.

  5. Web Tool: SUFE-CS-CONF-DDL, 2022.01-2022.03

    • Based on Vue-Cli, we built a countdown system tool for the computer conference of Shanghai University of Finance and Economics, which provides dual retrieval of tenure track Tier/CCF level.

    • Visiting the website: SUFE Tenure Track CS Conference Deadlines 🔍

  6. Analysis of Shanghai's biopharmaceutical industry chain based on big data and comparative study of the Yangtze River Delta, 2021.02-2022.01

    • Combining big data intelligent industrial research technology with traditional industrial economics, we constructed a knowledge graph of the biopharmaceutical industry chain and conducted an analysis of the biopharmaceutical industry chain in the Yangtze River Delta.

    • In May 2021, the leaders of the Shanghai Municipal Party Committee visited an economic regulatory platform in a certain district and fully recognized the achievements of the platform construction, pointing out that further efforts should be made to build a "city brain"🧠upgraded version in Shanghai.

  7. Policy and business classification in technology public opinion recommendation systems, 2020.08-2021.01

    • We crawled and cleaned policy texts released by the government, and conducted multi-label classification based on the business types of Ping An Technology's business group, with traditional and SOTA ML/DL text classification methods.

    • Partial code: Long-Text-Bert-Multi-label-Text-Classification-Pytorch

Education📔

Visiting Scholar, Business Analytics, Tippie College of Business, University of Iowa🦅, 2022.04-2023.04

PhD, Management Science and Engineering, School of Information Management & Engeineering, Shanghai University of Finance and Economics, 2018.09-2023.12

  • GPA: 3.6/4.0, Ranking: top10%

  • Main Courses: Machine Learning, Financial Engineering, Advanced Econometrics, Advanced Statistics, Advanced Operations Research, Optimization Theory, Stochastic Models.

  • Supervised by Professor Hailiang Huang and Songqiao Han

B.E., Statistics, College of Finance and Statistics, Hunan University, 2014.09-2018.06

  • GPA: 3.8/4.5, Ranking: top5%

  • Main Courses: Statistics, Multivariate Statistical Analysis, Sampling Techniques, Econometrics, Mathematical Statistics, Finance, Fundamentals of Programming, Statistical Software Applications.

Selected Competitions and Awards🏅

  1. National Scholarship, Ministry of Education of the People's Republic of China, 2017.11

  2. First-class Scholarship for Graduates, Shanghai University of Finance and Economics, 2022.12

  3. Second-class Scholarship for Graduates, Shanghai University of Finance and Economics, 2018.12&2019.12&2020.12&2021.12

  4. First-class Scholarship for Undergraduates, Hunan University, 2016.12

  5. Citibank Future Elite Scholarship, Citibank(China), 2016.12

  6. Outstanding Graduates of Hunan Province, Education Department of Hunan Province, 2018.06

  7. School of Merit student, Shanghai University of Finance and Economics, 2019.12

  8. School of Merit student, Hunan University, 2017.12

  9. School of Excellent Class Leader, Hunan University, 2015.12

  10. National First prize at the Agriculture Development Bank Cup Essay Competition, 2016.10

  11. National Second prize at the 7th National College Student Market Survey and Analysis Competition, 2017.05

  12. H Prize in the National Mathematical Contest in Modeling, 2016.03


Last Update: November, 2024.