Fwd: [CFP] IJCAI '24 Workshop on Recommender Systems in Finance, August 4 (Deadline May 4)
水田孝信 (Takanobu MIZUTA)
From: Yongjae Lee <yongj...@gmail.com>
Date: 2024年4月7日(日) 22:27
Subject: [aifin-worldwide] [CFP] IJCAI '24 Workshop on Recommender Systems in Finance, August 4 (Deadline May 4)
To: AI in Finance Worldwide <aifin-w...@googlegroups.com>
August 4, 2024, Jeju, Republic of Korea
Workshop website: https://sites.google.com/view/fin-recsys2024/
Submission deadline: May 4, 2024 11:59 pm (Anywhere on Earth)
Acceptance notification: June 4, 2023
The scope and topics of the proposed workshop are (broadly defined and not limited to) machine learning and optimization models for recommender systems in finance:
● Users
Retail investors, investment managers, financial advisors
● Items
Stocks, bonds, funds, loans, insurances, real estates
● Models
Collaborative filtering, item-based filtering, user-based filtering, sequential models, portfolio optimization, reinforcement learning, large language models (LLMs)
● Specific topics
Personalized recommendations, enhance investment performances, investor modeling, similarity learning of financial assets, explainability of investment recommendations, fairness and ethics, use of alternative data, evaluation of investment recommendations, generating synthetic datasets for investment recommendations, others (e.g., credit scoring, fraud detection)
Current state-of-the-art recommender systems utilize a vast amount of user-item interaction data to infer users’ preferences. Recommender systems are typically designed to account for user preferences(e.g., music or video streaming, online retail markets), but how to go about incorporating client preferences for financial recommender systems is not straightforward. As Mccreadie et al. (2021) points out, ‘the customer is not always right’ in investment management, because their preferences do not necessarily lead to good investment performances or feasible financial outcomes due to the volatility of markets. Therefore, investment recommender systems must consider the risk and return of a recommended asset or portfolio relative to the existing investor portfolio and preferences, as well as future predictions of the asset (Chung et al., 2023). Further, recommender systems in finance may not be targeted only at retail investors, but also finance professionals such as financial advisors who require new tools to oversee a broad array of clients, or investment managers who need fresh ideas to improve their portfolios.
Recommendations should be based on a mixture of investor demographics, financial situation and goals, and trading behaviors, and thus advanced investor modeling is paramount. Similarly, the similarity learning of financial assets (e.g., stocks, bonds, funds) is important for investment recommendations, because various features of financial assets (e.g., historical returns, future risk, factor exposure, financial statements, news) should be considered. On the other hand, the explainability of investment recommendations is crucial to persuade clients and avoid incomplete sales. Also, fairness and ethics of financial recommender systems need to be discussed.
Currently, there are many gaps between industry and academia in the area of financial recommender systems, where our previous workshops identified several specific issues:
1) Evaluation criteria: Good performance in terms of conventional recommendation performance measures (e.g., hit ratio, precision, recall, NDCG) does not guarantee good performance in terms of investment performance measures (e.g., return, volatility, Sharpe ratio), and vice versa.
2) Delivery method: There are many options for the delivery of financial recommendations, such as robo-advisors, hybrid advice from a financial-advisor-in-the-loop, interactive large language models (Lakkaraju et al., 2023), and more. The challenge within these models is incorporating financial goals into the investment strategy and updating advice often in a manner consistent with ensuring client uptake.
3) Nonstationarity of markets: Market dynamics significantly affect financial advice provided to clients and traders, where the mean and volatility of prices change over time. Recommender systems that are able to incorporate recent market events and respond to shock inputs are needed.
4) Lack of benchmark datasets: There is no public dataset for financial recommender systems, and thus, it is difficult to compare the performance of models proposed by different researchers. However, privacy issues make it challenging to share transaction data of retail investors in public and internationally. Hence, approaches to generating synthetic data for client-based data will be helpful in this regard.
This workshop will bring together academic and industry participants working on state-of-the-art research in designing financial recommender systems to discuss the topics mentioned above. This workshop builds on previous workshops on machine learning for investor modeling where we identified recommender systems as an emerging area of quantitative behavioral finance.
References:
● Chung, Munki, Yongjae Lee, and Woo Chang Kim. “Mean-Variance Efficient Collaborative Filtering for Stock Recommendation” arXiv preprint https://arxiv.org/abs/2306.06590
● McCreadie, R., Perakis, K., Srikrishna, M., Droukas, N., Pitsios, S., Prokopaki, G., ... & Ounis, I. (2021). Next-Generation Personalized Investment Recommendations. In Big Data and Artificial Intelligence in Digital Finance: Increasing Personalization and Trust in Digital Finance using Big Data and AI (pp. 171-198). Cham: Springer International Publishing.
● Lakkaraju, Kausik, Sara E. Jones, Sai Krishna Revanth Vuruma, Vishal Pallagani, Bharath C. Muppasani, and Biplav Srivastava. "LLMs for Financial Advisement: A Fairness and Efficacy Study in Personal Decision Making." In Proceedings of the Fourth ACM International Conference on AI in Finance (2023) pp. 100-107
Papers should be submitted on chartingtool by May 4th, 2024.
Submission link: https://chairingtool.com/conferences/FINRECSYS24/MainTrack
All submissions should follow IJCAI guidelines (https://ijcai24.org/call-for-papers/), except for page limit and submission dates.
Two types of submissions are accepted:
short papers 4-8 pages (excluding references)
extended abstracts 2 pages (excluding references)
Accepted papers require in-person presentation by at least one author. All accepted papers will be published on the workshop website, and authors are encouraged to also share their work on platforms like arXiv or other online repositories.
Paper submission deadline: May 4, 2024
Author notification: June 4, 2024
Camera ready version: July 4, 2024
Workshop: August 4, 2024
Additionally, three members of the organizing committee are guest editors for the special issue on “Statistical and Machine Learning for Investor Modelling” at the Journal of Behavioral Finance. The authors of accepted workshop paper submissions will be encouraged to submit full manuscripts to the special issue.
Link: https://think.taylorandfrancis.com/special_issues/statistical-machine-learning-investor-modelling/
- Yongjae Lee, UNIST, South Korea (lead organizer, email: yongj...@unist.ac.kr)
- John R.J. Thompson, University of British Columbia, Canada
- Dhagash Mehta, BlackRock, USA
- Thomas De Luca, Vanguard, USA
- Richard Mccreadie, University of Glasgow, UK
- Jaesik Choi, KAIST, South Korea
- Min Hee Kim, Hana Institute of Technology, South Korea
You received this message because you are subscribed to the Google Groups "AI in Finance Worldwide" group.
To unsubscribe from this group and stop receiving emails from it, send an email to aifin-worldwi...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/aifin-worldwide/f7073150-1589-4d55-9ebc-b7c2bfa33b71n%40googlegroups.com.