project-proposal-2025

[QTrade - Quantitative Trading AI Platform]

Quantitative Trading + AI Model Training

Abstract

With the advancement of Artificial Intelligence (AI) and big data analytics, an increasing number of traders and institutions are leveraging quantitative trading to optimize investment strategies. However, most existing quantitative trading tools rely on pre-defined models, which lack flexibility in adapting to users’ unique trading philosophies. Additionally, training and deploying a customized AI trading model requires a specialized data science team, making it unaffordable for individual investors and small to medium-sized institutions.

QTrade (Quantitative Trade) aims to solve this problem. It is an AI-powered quantitative trading training and execution platform that allows users to train, optimize, and deploy personalized AI models based on their trading philosophies and apply them directly to live market trading. The platform provides an intuitive GUI interface, enabling users to build trading strategies with no-code or low-code tools, train them into AI models, and use them for automated trading.

Author

Name: Haotian Xue

Student number: 48244990

Functionality

General Overview QTrade offers an intuitive web-based application that allows users to build and train their own AI trading models without requiring programming expertise. Users can utilize historical data and real-time market data to design their trading strategies based on technical analysis, fundamental analysis, and market sentiment analysis, and train a customized AI model for their needs.

Workflow User Login & Model Selection

After logging in, users can browse a marketplace of pre-trained AI models, each with a description of its strategy, past performance, and use cases. Selecting Data Sources

Upload personal historical trading data Use built-in financial databases (stocks, forex, futures, cryptocurrencies, etc.) Connect to trading platforms via API (e.g., Binance, NASDAQ, NYSE, Coinbase) Strategy Development Users can create trading strategies in multiple ways:

Select a pre-trained AI model and fine-tune it Design a new trading strategy via GUI (e.g., candlestick analysis, moving averages, momentum trading, Martingale strategy) Use Python/Solidity to define advanced custom trading strategies Adjusting Model Parameters

Users can fine-tune risk management, trade frequency, position sizing, stop-loss settings, and other parameters. Training the AI Model

The model is trained on historical data and automatically optimizes its parameters. Users can evaluate model performance based on profitability, win rate, maximum drawdown, Sharpe ratio, etc. Deploying the AI Trading Model

Cloud-based deployment: Users can host the model on QTrade’s cloud servers and connect it to their real trading accounts for automated execution. Local deployment: Users can download the trained model and run it on other trading platforms.

Supported Model Types QTrade will support the following trading models and provide a database to store users’ historical trading data and trained AI models:

Technical Analysis Strategies Fundamental Analysis Models Market Sentiment Analysis High-Frequency Trading (HFT) Strategies Machine Learning/Deep Learning Quant Models (Reinforcement Learning, LSTM, Transformer-based models) Hybrid Strategies (combining multiple methods to optimize trading decisions)

Scope

core functionalities:

  1. Enable users to train AI trading models online (via GUI or Python code)
  2. Allow users to import historical data or fetch real-time data via API
  3. Provide AI model backtesting functionality (visualizing performance metrics such as equity curve, risk management, and win-loss ratio)
  4. Enable AI trading strategy hosting and deployment (connect AI models to real trading accounts for automated execution)
  5. Offer API support so users can integrate trained models into third-party trading systems

Quality Attributes

  1. Modularity (Most Important) The trading strategy system must be highly flexible, allowing users to freely combine different data processing methods, model types, and execution logic. Decoupling is crucial for components such as data ingestion, model training, and trade execution to support future expansion and integrations. Example: Different trading APIs (Binance, NASDAQ, Coinbase) may have varying data formats, so the data processing module must provide standardized interfaces.
  2. Scalability Quantitative trading involves large-scale data processing (historical candlestick data, market sentiment analysis, etc.), requiring distributed computing and cloud deployment. AI model training is computationally intensive, so the platform must support vertical and horizontal scaling to prevent training bottlenecks. High-frequency trading (HFT) requires a low-latency execution engine, ensuring seamless operation with multiple simultaneous users.
  3. Extensibility The financial market evolves rapidly, requiring quick integration of new trading strategies, data sources, and trading APIs. The system architecture should follow a plugin-based design, allowing easy expansion. Example: If users want to add a Transformer-based AI trading model, it should seamlessly integrate into the existing system.
  4. Security Since real financial transactions are involved, data security and account security are paramount. The system should support Two-Factor Authentication (2FA) to prevent unauthorized account access. API access control must be enforced to ensure only authorized users can execute trades.

Evaluation

  1. Modularity: Assess code decoupling, ensuring each module functions independently and is interchangeable.
  2. Scalability: Benchmark AI training times and ensure GPU computing supports horizontal scaling.
  3. Extensibility: Verify the plugin-based architecture by integrating a new trading API or model seamlessly.
  4. Security: Conduct penetration testing to ensure that account and trading data remain secure.