// the find
charliedream1/ai_quant_trade
股票AI操盘手：从学习、模拟到实盘，一站式平台。包含股票知识、策略实例、大模型、因子挖掘、传统策略、机器学习、深度学习、强化学习、图网络、高频交易、C++部署和聚宽实例代码等，可以方便学习、模拟及实盘交易
A Chinese-language educational monorepo for AI-driven quantitative trading, covering everything from dual moving average basics to RL agents, LLM fine-tuning for stock prediction, and C++ deployment. It's aimed at Chinese retail investors and finance students who want worked examples across the full quant stack, not a production-ready system. Think of it as a well-organized notebook collection with commentary.
The breadth of strategy coverage is genuinely useful — you get RL (FinRL/StableBaselines3), classical factor libraries (alpha101, ta-lib), LLM fine-tuning with Unsloth, and graph neural networks all in one place with actual runnable notebooks. The resource curation section (a_全网优秀资源) is the real gem: annotated comparisons of backtest frameworks, factor libraries, and LLM-finance papers that would take days to assemble on your own. Active maintenance through mid-2026, with the inference-model stock prediction tutorial added recently and a reported 20% accuracy improvement over base models. The Jupyter-first format means you can actually trace what's happening in each strategy without fighting a framework abstraction.
Almost nothing here is production-ready — the live trading code depends on Wind (expensive institutional data terminal) and the RL backtests have no walk-forward validation or realistic transaction cost modeling, so the 53% annualized return figure is essentially meaningless without that context. The repo is a notebook dump, not a library: no shared interfaces, no consistent data pipeline, and the requirements.txt is a single flat file that will conflict depending on which subdirectory you're actually running. Most of the LLM trading content (FinGPT, TradingAgents, etc.) is pointer documentation to other projects, not implemented code — you'll follow links to separate repos that may have diverged. Non-Chinese-speaking users are second-class citizens despite the English README existing; most tutorial content, comments, and the paid 'knowledge planet' community are Chinese-only.