QUANTUMFLOW: REAL-TIME ALGORITHMIC TRADING ENGINE WITH LSTM, MONTE CARLO SIMULATION, AND TECHNICAL INDICATORS DEPLOYED ON EMBEDDED HARDWARE

Abstract

This invention relates to a real-time algorithmic trading system deployed on Raspberry Pi, designed for executing trades in Futures and Options (F&O) markets. The system integrates machine learning models (LSTM networks) with technical indicators (EMA and RSI) and Monte Carlo simulations to ensure high prediction accuracy and robust risk management. The LSTM model analyzes historical stock data to predict price movements, while the EMA and RSI strategies confirm trade signals. The Monte Carlo simulation evaluates the risk of each trade, and only trades with acceptable risk levels are executed. The decision-making unit consolidates these inputs, placing trades via Upstox APIs when the combined confidence level exceeds 68%. The system's deployment on Raspberry Pi offers a lightweight, portable, and cost-effective trading solution with low latency. This invention provides an accessible platform for individual traders, ensuring optimized trade frequency, minimized risks, and enhanced profitability.

Specification

FORM 2
THE PATENT ACT 1970
(39 of 1970)
The Patents Rules, 2003 COMPLETE SPECIFICATION
(See section 10 and rule 13)
1. TITLE OF THE INVENTION - QuantumFlow: Real-Time Algorithmic Trading Engine with LSTM, Monte Carlo Simulation, and Technical Indicators Deployed on Embedded Hardware
2. APPLICANT (S)
1. (a) NAME: Surabhi Sanjiv Kakade
(b) NATIONALITY: Indian
(c)ADDRESS: Vishwakarma Institute of Technology, 666 Upper Indira Nagar, Bibwewadi, Pune-411037
2. (a) NAME: Tanaya Prashant Korhalkar
(b) NATIONALITY: Indian
(c)ADDRESS: Vishwakarma Institute of Technology, 666 Upper Indira Nagar, Bibwewadi, Pune-411037
3. (a) NAME: Bhavesh Nitin Patil
(b) NATIONALITY: Indian
(c) ADDRESS: Vishwakarma Institute of Technology, 666 Upper Indira Nagar, Bibwewadi,
Pune-411037
4. (a) NAME: Heramb Sachin Bhoodhar
(b) NATIONALITY: Indian
(c) ADDRESS: Vishwakarma Institute of Technology, 666 Upper Indira Nagar, Bibwewadi, Pune-411037
5. (a) NAME: Amey SudhirPathe
(b) NATIONALITY: Indian
(c) ADDRESS: Vishwakarma Institute of Technology, 666 Upper Indira Nagar,
Bibwewadi, Pune-411037
6. (a) NAME : Chetana Pitambar Ingale
(b) NATIONALITY: Indian
(c) ADDRESS: Vishwakarma Institute of Technology, 666 Upper Indira Nagar, Bibwewadi,
Pune-411037

3.PREAMBLE OF THE DESCRIPTION
This invention relates to the development of a real-time algorithmic trading system, specifically designed for automated trading of financial instruments like Futures and Options (F&O). The invention combines machine learning models, statistical simulations, and financial indicators to achieve accurate predictions and risk-aware trading decisions. The system is deployed on a Raspberry Pi, ensuring portability, low cost, and energy efficiency, suitable for small-scale traders and individual investors.

4. Description 4.1 Field of the Invention
This invention falls within the fields of financial technology, algorithmic trading, and predictive analytics. It introduces a trading engine capable of executing trades in real-time for financial instruments such as Nifty and Bank Nifty Futures and Options. The system integrates a Long Short-Term Memory (LSTM) model for stock price prediction, Monte Carlo simulations for risk analysis, and technical indicators like Exponential Moving Average (EMA) and Relative Strength Index (RSI) for trade confirmation. Unlike traditional high-end trading platforms, the invention runs on a Raspberry Pi, offering a low-cost, portable, and accessible solution.
4.2 Background of the Invention
Algorithmic trading systems have become essential in modern financial markets, allowing for automated trade execution and data-driven decision-making. Traditional trading models rely heavily on technical indicators to detect trends and predict stock price movements, but these indicators often struggle with highly volatile and non-linear financial markets. The emergence of machine learning models like LSTM networks has improved the forecasting of stock prices by capturing long-term dependencies in time-series data.
However, the existing systems used for financial forecasting often require high-performance computing resources, limiting their accessibility. Additionally, risk management remains a critical concern, as high-frequency trades can expose traders to excessive risks. This invention addresses these challenges by combining machine learning predictions, risk simulations, and technical analysis in a single platform deployed on Raspberry Pi hardware. The result is a low-cost, portable trading engine that delivers accurate predictions while mitigating risks in real¬time.
4.3 Objectives of the Invention
1. Accurate Stock Price Prediction: To provide high-confidence predictions using LSTMmodels trained on historical stock data.
2. Seamless Risk Management: To integrate Monte Carlo simulations for quantifying therisks associated with each trade before execution.
3. Trade Confirmation Using Technical Indicators: To enhance decision-making throughEMA and RSI-based strategies, ensuring optimal entry and exit points.
4. Real-Time Execution in F&O Markets: To fetch live market data and place orders inNifty and Bank Nifty using Upstox APIs and WebSocket connections.
5. Low-Cost Hardware Deployment: To deploy the trading engine on a Raspberry Pi,making it affordable and accessible for small traders and individual investors.
6. Latency Reduction: To minimize data-fetching delays and trade execution time throughlightweight code and optimized API requests.

4.4 Advantages of Invention
1. Portability and Accessibility: The use of Raspberry Pi allows the system to be deployedanywhere, making it ideal for remote traders or on-the-go operations.
2. Cost Efficiency: Unlike conventional trading systems requiring expensive servers or high-end computers, this invention provides an affordable solution without compromising on functionality.
3. Accurate Predictions: The integration of LSTM networks improves forecasting accuracy by identifying long-term dependencies and reducing the effect of market noise.
4. Reduced Trading Risks: The Monte Carlo simulations assess the probability of losses before trade execution, ensuring only low-risk trades are pursued.
5. Optimized Trade Frequency: With the 68% confidence threshold for decision¬making, the system ensures fewer but more accurate trades, balancing risk and reward.
6. Real-Time Performance: The combination of Upstox APIs and WebSocket connectionsensures instant data fetching and trade execution with minimal latency.
4.5 Summary of the Invention
The invention provides a real-time, low-cost algorithmic trading engine capable of predicting stock price movements, evaluating risks, and executing trades. It integrates machine learning techniques (LSTM) with Monte Carlo simulations and technical indicators (EMA and RSI) to deliver optimized trade strategies. Deployed on a Raspberry Pi, the system ensures portability and cost-efficiency while offering high performance. Live market data is fetched through Upstox APIs, and trades are executed only when the system achieves a minimum 68% confidence level based on combined model predictions.

4.6 Brief Description of the drawing:
Fig 1: System Architecture of QuantumFlow
The system architecture of the proposed algorithmic trading engine illustrates the seamless interaction between the key hardware and software components deployed on a Raspberry Pi. At the core of the system, the Raspberry Pi acts as the primary computational device, running the backend logic through Python Flask. The hardware is connected to Upstox APIs and WebSocket services, which allow the system to fetch live market data and execute trades with minimal latency. The data flows into different modules, each responsible for specific tasks such as stock price prediction, risk assessment, and trade validation.
The LSTM model forms the heart of the predictive engine, processing historical stock data to generate probabilities for Put and Call trades. This prediction is further enhanced by the custom technical strategy that computes Exponential Moving Averages (EMA) across 9, 20, and 50-day periods, along with Relative Strength Index (RSI) values within the 40-60 range to confirm market trends. These modules interact with the Monte Carlo simulation engine, which analyzes multiple scenarios to determine the potential risk of executing the predicted trade A decision making unit consolidates the outputs of the LSTM, EM A/RSI strategy, and

Monte Carlo simulation. Only when the confidence level surpasses a predefined threshold (68%) and the risk analysis approves the trade, the system places an order in the live market through the Upstox API.
The diagram depicts the flow of data from real-time data sources to prediction and decision¬making modules, leading to trade execution. This modular structure ensures that each component works cohesively, balancing prediction accuracy with risk management to ensure optimal trade performance.
4.7 Detailed Description of the drawing:
The detailed architecture of the system highlights the functional components and their interactions. At its core, the Raspberry Pi serves as a lightweight but powerful computing platform, enabling the deployment of the trading engine in a cost-effective and portable manner. The Raspberry Pi runs a Python Flask backend, which orchestrates communication between the data sources, prediction models, and trade execution modules.
The LSTM model is responsible for generating trade signals by analyzing historical stock data. The data is retrieved either from yfinance or through Upstox APIs, covering key financial metrics such as open, close, high, and low prices. The model outputs confidence scores for both Call and Put trades. For example, the system may predict a 67% confidence for a Call trade and 33% for a Put trade. This output is further processed by the EMA/RSI strategy module, which uses technical indicators to validate trends. The EMA is computed over 9, 20, and 50-day periods to identify moving trends, while the RSI ensures that the market is not in an overbought or oversold state.
The Monte Carlo simulation module acts as the risk assessment engine, simulating thousands of potential market scenarios. It evaluates the probability of losses for each trade decision, providing a ! statistical safeguard. If the Monte Carlo simulation identifies an unacceptable risk, the system avoids executing the trade, even if the prediction confidence is high. This layer ensures that the system prioritizes risk management over frequent trading.
All the outputs from the LSTM model, EMA/RSI strategy, and Monte Carlo simulation are fed into the decision-making unit, which determines whether to execute a trade. The system is designed to trigger a trade only if the combined confidence from the LSTM and technical strategy exceeds 68%. This threshold ensures that only high-confidence trades are executed, balancing accuracy with risk mitigation.
The Upstox API module is responsible for both fetching live market data and placing trades in real-time. The API integration enables the system to monitor market conditions continuously and place orders instantly when the decision-making unit signals a trade. To minimize latency, the system relies on WebSocket connections that provide real-time updates, ensuring that trades are executed without delay.
This architecture ensures that every component plays a vital role in optimizing trade performance. The Raspberry Pi provides a compact and energy-efficient platform for deploying the entire trading engine, making it highly accessible for small traders and individual investors. By integrating machine learning predictions, technical indicators, and risk simulations, the system achieves a balance between profitability and safety. The modular design also allows for future enhancements, such as adding new technical indicators or improving model performance through updated training datasets. This interconnected structure ensures that the trading engine remains scalable, efficient, and reliable, even under volatile market conditions

5. CLAIMS
I/We claim that
l.A real-time algorithmic trading system deployed on a Raspberry Pi, integrating machine learning models, financial indicators, and risk management strategies for tradeexecution in Futures and Options (F&O) markets.
2. The system employs a Long Short-Term Memory (LSTM) model to predict stock
price movements by analyzing historical market data, generating confidence scores for potential trades.
3. A custom strategy using Exponential Moving Average (EMA) and Relative Strength
Index (RSI) confirms market trends and enhances the reliability of trade decisions.
4. Monte Carlo simulations are incorporated to assess trade risks by evaluating multiple
market scenarios, allowing only low-risk trades to proceed.
5. A decision-making unit integrates outputs from the LSTM model, EMA/RSI
strategy, and Monte Carlo simulation to trigger trade execution through Upstox
APIs, ensuring trades are placed only when the combined confidence surpasses
68%.

Documents