init

Author: TobyYang7

.gitignore

@@ -0,0 +1 @@
+.DS_Store

readme.md

@@ -0,0 +1,56 @@
+# TwinMarket: A Scalable Behavioral and Social Simulation for Financial Markets
+
+<img src="./src/TwinMarket.jpg" alt="Logo" >
+
+## Overview
+**TwinMarket** is a novel multi-agent framework designed to simulate socio-economic systems using large language models (LLMs). The framework focuses on modeling individual investor behaviors and their interactions within a simulated stock market environment. By leveraging the **Belief-Desire-Intention (BDI)** framework, TwinMarket captures the cognitive processes of agents, enabling the study of emergent phenomena such as financial bubbles, recessions, and market volatility.
+
+The project aims to bridge the gap between micro-level individual decision-making and macro-level collective market dynamics, providing insights into how individual actions aggregate to form complex socio-economic patterns.
+
+## Key Features
+1. **Real-World Alignment**: The framework is grounded in established behavioral theories and calibrated with real-world data, ensuring realistic human behavior modeling.
+2. **Dynamic Interaction Modeling**: TwinMarket captures diverse human behaviors and their interactions, particularly in the context of information propagation and social influence.
+3. **Scalable Market Simulations**: The framework supports large-scale simulations, allowing researchers to analyze the impact of group size and interaction complexity on market behavior.
+
+## Framework Components
+### Micro-Level Simulation: Individual Behaviors
+- **BDI Framework**: Agents are modeled using the **Belief-Desire-Intention** framework, which structures their decision-making processes.
+- **Behavioral Biases**: Agents exhibit various behavioral biases such as overconfidence, loss aversion, and herding behavior, reflecting real-world investor psychology.
+
+### Macro-Level Simulation: Social Interactions
+- **Social Network Construction**: Agents interact within a dynamic social network, where connections are based on trading behavior similarity.
+- **Information Propagation**: The framework models how information spreads through the network, leading to phenomena like opinion polarization and echo chambers.
+
+### Data Sources
+
+![1738736488423](./src/data.png)
+
+- **Real-World Data**: TwinMarket integrates real user profiles, transaction details, stock data, and news articles to create a realistic simulation environment.
+- **Initial User Profiles**: User profiles are generated using real transaction data from platforms like Xueqiu, ensuring diversity in agent behavior.
+
+## Experimental Results
+TwinMarket successfully replicates key stylized facts of financial markets, including:
+- **Fat-tailed return distributions**
+- **Volatility clustering**
+- **Leverage effects**
+- **Volume-return relationships**
+
+The framework also demonstrates the emergence of group behaviors, such as self-fulfilling prophecies and information cascades, which are difficult to capture using traditional agent-based models.
+
+## Scalability
+TwinMarket is designed to scale to large populations, with simulations involving up to 1,000 agents. The framework maintains realistic market dynamics even at larger scales, providing a robust platform for studying complex socio-economic systems.
+
+## How to Cite
+If you use TwinMarket in your research, please cite the following paper:
+
+```bibtex
+@misc{yang2025twinmarketscalablebehavioralsocialsimulation,
+      title={TwinMarket: A Scalable Behavioral and SocialSimulation for Financial Markets}, 
+      author={Yuzhe Yang and Yifei Zhang and Minghao Wu and Kaidi Zhang and Yunmiao Zhang and Honghai Yu and Yan Hu and Benyou Wang},
+      year={2025},
+      eprint={2502.01506},
+      archivePrefix={arXiv},
+      primaryClass={cs.CE},
+      url={https://arxiv.org/abs/2502.01506}, 
+}
+```