Millennium Project — Quantitative Trading & Portfolio Simulation Platform

A full-stack quantitative trading simulation system combining order management, portfolio tracking, option analytics, predictive modeling, and AI-powered trade advice.

This project replicates the core functions of an institutional trading system — from order routing and execution to forecasting and strategy recommendations — built entirely with Python, Flask, MySQL, and machine learning.

---

Project Overview

The Millennium Project simulates a miniature trading platform that integrates:

1. Order Management & Execution Engine
2. Portfolio Tracking & P&L Analysis
3. Options Pricing & Volatility Surface Visualization
4. Predictive Modeling using LSTM Networks
5. AI-driven Order Parsing and Trade Advice (Gemini API)

---

Tech Stack

Backend: Flask (Python), SQLAlchemy, MySQL
Frontend: HTML, CSS, Chart.js
Data APIs: Yahoo Finance (yfinance)
Machine Learning: TensorFlow/Keras (LSTM), pandas-ta
AI Integration: Gemini API (Natural Language Processing)

---

System Architecture & Data Flow

User Input → Flask Backend → MySQL Database → Execution Engine ↘ Gemini NLP → Order Parsing ↘ Yahoo Finance → Market Data, Options, Forecasts ↘ LSTM Model → Price Prediction ↘ Chart.js Frontend → Visualization

Workflow Summary:

1. User places an order (manual or via AI command).
2. Flask backend validates and routes order.
3. Execution engine simulates fills based on live market data.
4. Portfolio and P&L values update automatically.
5. LSTM forecasts future price movement.
6. Gemini API generates actionable trading advice.

---

1. Order Management & Execution

Finance Context

In financial markets, orders pass through an Order Management System (OMS) to be routed to an exchange or broker.
The system replicates this flow by applying execution rules against live quotes.

Order Types Supported:

• Market Order: Execute immediately at best available price.
• Limit Order: Execute only at or better than a specified price.
• Stop Order: Converts to market order when stop price triggered.
• Stop-Limit Order: Becomes a limit order once stop threshold is reached.

Technical Implementation

• Flask REST API endpoint /orders accepts POST requests with JSON payloads.
• MySQL Schema (SQLAlchemy):

id | symbol | side | qty | type | limit_price | stop_price | status | filled_price | timestamp

• Execution Logic:
• Market → Filled at current quote.
• Limit → Filled only when bid/ask crosses.
• Stop → Triggered when price hits stop threshold.

Simulates a real exchange matching engine locally.

---

2. Portfolio Management & P&L Tracking

Finance Context

A portfolio tracks positions in each security:

Unrealized P&L = (Market Price − Cost Basis) × Quantity
Realized P&L = Gains/losses from closed trades

Technical Implementation

• Positions table updated dynamically after order execution.
• Valuation loop: Fetches real-time market data via yfinance.
• Portfolio performance chart: Aggregates portfolio value over the last 30 days using Chart.js.

Core features include cost basis, market value, and total portfolio performance visualization.

---

3. Options Pricing & Volatility Surface

Finance Context

Options derive value from:

• Underlying stock price
• Strike price
• Expiration
• Interest rate
• Volatility (σ)

Black-Scholes Model (BSM):
[ C = S_0 N(d_1) - Ke^{-rT} N(d_2) ] where implied volatility (IV) is the σ that equates market price with model price.

Technical Implementation

• Pulled option chains using yfinance.
• Inverted Black-Scholes with Newton’s method to find IV.
• Plotted 3D volatility surface (strike × expiration × IV) with Chart.js.
• Built option payoff diagrams for strategy visualization.

Visual analytics replicate institutional derivatives dashboards.

---

4. Predictive Modeling (LSTM Network)

Finance Context

Stock prices form time series — exhibiting autocorrelation, momentum, and trend.
LSTM networks model sequential dependencies better than traditional regressions.

Technical Implementation

• Pulled 6 months of OHLCV data via yfinance.
• Generated technical indicators (SMA, EMA, RSI) with pandas-ta.
• Scaled features using MinMaxScaler.
• Built LSTM model:
• Input (timesteps × features)
• LSTM hidden layers
• Dense output predicting next-day price
• Produced 7-day forecast curve and classified market trend as bullish/bearish/neutral.
• Cached trained models for efficiency.

Forecasts guide strategy signals — not as sole trading decisions, reflecting real quant practices.

---

5. AI Order Parsing & Trading Advice

Finance Context

Traders often issue natural language orders ("Buy 10 AAPL at 150 if it dips").
Institutions automate this via NLP and voice-trade interfaces.

Technical Implementation

• Gemini API Integration:
• Input: Natural language command
• Output: JSON structure (symbol, qty, type, limit_price)
• Trading Advisor:
• Combines LSTM forecast + portfolio state
• Prompts Gemini to generate short strategy advice

Example Output:

“AAPL forecast shows +3% over 7 days. Hold long; consider stop at $142.”

Mimics sell-side research desks delivering concise recommendations.

---

6. Full System Integration

Data Sources:

• Yahoo Finance (quotes, options, history)
• Gemini API (NLP + trade advice)
• MySQL (persistent portfolio + orders)

Frontend Visualizations:

• Portfolio performance curve
• Option payoff diagrams
• Volatility surface
• Forecast trends

All analytics rendered dynamically using Chart.js.

---

Tools & Libraries

Category	Libraries / Tools
Backend	Flask, SQLAlchemy, MySQL
Market Data	yfinance
Machine Learning	TensorFlow / Keras, pandas-ta
Frontend	HTML, CSS, Chart.js
AI Integration	Gemini API
Data Processing	pandas, numpy

---