Stock Trading Order Execution via Reinforcement Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional algorithmic trading systems lack agility in responding to stock market changes, leading to inefficiencies when executing large orders, particularly in terms of market adaptability and cost optimization.
Innovation Solution
A reinforcement learning-based deep learning model is employed to derive and execute a strategy for stock orders in real-time, utilizing pre-trained models for supervised and reinforcement learning to optimize order execution based on trading data, minimizing losses and slippage costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional algorithmic trading systems execute orders in parts according to predetermined rules, then order execution is systematic and automated, but the system lacks agility to respond appropriately to changes in stock market conditions
Solution Approach 1:
The patent implements a feedback mechanism where the trading system continuously monitors market conditions and adjusts order execution parameters in real-time. The system receives feedback from market data feeds, analyzes changes in stock price trends, volume patterns, and other market indicators, then dynamically modifies execution strategies to respond appropriately to market movements, resolving the contradiction between systematic execution and market adaptability
Solution Approach 2:
The patent transitions from static predetermined rules to dynamic adaptive algorithms that can change parameters during execution. The system uses real-time market data to adjust order size, timing, and pricing strategies dynamically, allowing the trading system to adapt to changing market conditions while maintaining automated execution, thus balancing adaptability with system complexity
2Productivity
If a time-weighted average price (TWAP) algorithm executes an order in equally divided parts as time passes, then execution is simple and automated, but the system has low market adaptability and low efficiency
Solution Approach 1:
The patent changes the parameters of order execution from fixed equal divisions to dynamic variable divisions based on market conditions. The system adjusts parameters such as order size, time intervals, and pricing strategies in real-time based on market volatility, liquidity conditions, and price trends, thereby improving both trading efficiency and market adaptability simultaneously
Solution Approach 2:
The patent incorporates preliminary analysis of market conditions before executing orders. The system pre-processes market data, identifies optimal execution windows, and prepares adaptive strategies in advance based on predicted market movements, allowing more efficient and adaptable order execution without requiring complex real-time adjustments during critical trading moments
3Quantity of substance
If large orders are executed in the stock market, then trading volume is achieved, but the stock price may sharply fall or rise making it difficult to trade stocks at a right price
Solution Approach 1:
The patent applies segmentation by dividing large orders into multiple smaller child orders that are executed sequentially or in parallel based on market conditions. This segmentation reduces the market impact of each individual order execution, minimizing price slippage while still achieving the overall trading volume objective. The system dynamically adjusts the segmentation strategy based on real-time market liquidity and volatility
Solution Approach 2:
The patent introduces an intermediary algorithmic trading system that acts as a mediator between the large order and the market. This intermediary system uses real-time market data to intelligently distribute orders across different time periods and price levels, smoothing out the impact on market price and reducing slippage costs while achieving the desired trading volume
Data Source
AI summary
An order execution server for stock trading includes a data collection unit configured to collect trading data on at least one item, a subsidiary prediction value generation unit configured to generate a subsidiary prediction value by inputting the trading data into a pre-trained first deep learning model based on supervised learning, an order execution strategy deriving unit configured to derive an order execution strategy for the at least one item during a current period of time based on the trading data and the subsidiary prediction value by using a pre-trained second deep learning model based on reinforcement learning; and an order execution instruction unit configured to instruct order execution for the at least one item during the current period of time by using order information including the order execution strategy.