A high-frequency trading system based on artificial intelligence

Abstract

The present invention discloses a high-frequency trading system based on artificial intelligence, including a global trading market aggregation management system, a quantitative arbitrage trading hedging system and a risk control management system; the output end of the global trading market aggregation management system is connected to a market processing and analysis subsystem, the output end of the market processing and analysis subsystem is connected to an order queue management system, and the output end of the order queue management system is connected to a number of arbitrage trading modules. The high-frequency trading system can better realize the functions of enhancing market liquidity, improving price stability, preventing price manipulation, and market price discovery. In addition, the system takes computer vision algorithm as the core, uses market order queue data analysis and monitoring middleware as the "eyes", and constructs a decision-making system for monitoring abnormal transactions and order changes.

Description

Technical Field

[0001] The present invention relates to the technical field of high-frequency trading, and in particular to a high-frequency trading system based on artificial intelligence. Background Art

[0002] Commodities refer to physical goods that can enter the circulation field but are not in the retail stage, have commodity attributes and are used in industrial and agricultural production and consumption in large quantities. In the financial investment market, commodities refer to homogeneous, tradable commodities that are widely used as industrial basic raw materials, such as crude oil, non-ferrous metals, steel, agricultural products, iron ore, coal, etc. It includes three categories, namely energy commodities, basic raw materials and agricultural and sideline products.

[0003] Currently, commodities have a serious liquidity scarcity problem. For traders, this is specifically reflected in: the market size is small, most institutions and users have not entered the industry, the trading market is relatively small, traffic is dispersed, liquidity costs are high, large transactions are difficult, and transaction risks are high. In order to achieve functions such as enhancing market liquidity, improving price stability, preventing price manipulation, and market price discovery, a trading system that can aggregate liquidity across the entire industry is needed. Summary of the invention

[0004] The purpose of the present invention is to provide a high-frequency trading system based on artificial intelligence to solve the problems raised in the above background technology.

[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a high-frequency trading system based on artificial intelligence includes a global trading market aggregation management system, a quantitative arbitrage trading hedging system and a risk control management system;

[0006] The output end of the global trading market market aggregation management system is connected to a market processing and analysis subsystem, the output end of the market processing and analysis subsystem is connected to an order queue management system, and the output end of the order queue management system is connected to a plurality of arbitrage trading modules;

[0007] The output end of the quantitative arbitrage trading hedging system is connected to a core trading queue algorithm control matching engine pool module, and the output end of the core trading queue algorithm control matching engine pool module is connected to a reverse hedging control engine module;

[0008] The output end of the risk control management system is connected to an account net value and market matching control module, the output end of the account net value and market matching control module is connected to a transaction order management system, and the output end of the transaction order management system is connected to several hedging transaction market modules;

[0009] The output end of the order queue management system is connected to the core transaction queue algorithm control matching engine pool module, the output end of the account net value and market matching control module is connected to the core transaction queue algorithm control matching engine pool module, and the output end of the reverse hedging control engine module is connected to the transaction order management system.

[0010] Furthermore, the global trading market quotation aggregation management system includes an extremely fast quotation acquisition module, a monitoring middleware module and an aggregation management module. The extremely fast quotation acquisition module is used to acquire quotation and data, and the quotation and data include historical data, real-time quotation, factor data, tick data and market data. The monitoring middleware module is used to monitor market fluctuations, price spreads, cross-exchange monitoring and background data monitoring. The aggregation management module is used to aggregate market orders from multiple exchanges.

[0011] Furthermore, the market processing and analysis subsystem can perform extremely fast market analysis, fully extract market characteristics, and improve the algorithm's anti-noise ability based on algorithms and deep learning.

[0012] Furthermore, the order queue management system includes an order identification module and a data analysis module. The order identification module is used to perform order queue identification, abnormal order identification, active buy order volume and sell order volume identification, and large transaction order feature identification. The data analysis module can predict queue orders based on the market quotation of the machine learning algorithm. The order queue control and quotation determine the market making arbitrage ability. The strategy is highly sensitive to the system response speed and delay, and captures price fluctuations near the market bid and ask prices in a very short time through high-speed buying and selling.

[0013] Furthermore, the arbitrage trading module includes an order execution module and a strategy module. The order execution module is used to execute pending orders, buried orders, orders and order cancellations. The strategy module captures the price difference between two financial assets with exactly the same underlying assets to obtain profits through arbitrage strategies, high-frequency trading strategies and neutral strategies.

[0014] Furthermore, the quantitative arbitrage trading hedging system includes a trading front-end module, a user module and a network management module. The trading front-end module is used to realize the functions of monitoring transactions and customer link management. The user module can generate orders with quantity commissioned by terminal users and send them to the trading platform. The network management module is used to collect user orders and distribute the collected user orders to the core trading queue algorithm to control the matching engine pool module.

[0015] Furthermore, the core transaction queue algorithm controls the matching engine pool module, including a matching processing module, a transaction record log module and a memory data module. The matching processing module is the core part of the transaction system, which is used to receive orders and implement order matching according to business logic, generate transaction records at the same time, and then give users transaction result feedback. The transaction record log module can record orders and transactions, and the memory data module is used to store orders and transaction records during the transaction process to achieve data persistence.

[0016] Furthermore, the reverse hedging control engine module can control its own risk exposure through hedging strategies, wherein the hedging strategy will control the risk exposure through the Greek letter values ​​of delta, gamma, vega, and thetahe rho. The most basic principle is to keep delta in the middle to avoid the impact of target price fluctuations on the system's positions. After the researcher completes the research on the hedging strategy, the program manager will implement the hedging strategy in the system. During the transaction, the system can automatically hedge according to the transaction situation, select other pricing contracts or target spot according to the hedging strategy model, and construct arbitrage combinations, pricing combinations, and pricing and spot combinations to achieve inventory risk control.

[0017] Furthermore, the risk control management system can perform data monitoring, data analysis, exception handling and model iteration based on algorithms, and conduct risk control, monitoring model risk, liquidity risk, operational risk, information asymmetry risk and survival risk before, during and after the event.

[0018] Furthermore, the account net value and market matching control module is used to match the account net value and market conditions for trading, the transaction order management system module manages transaction order information, and the hedging transaction market module performs hedging when the strategy is triggered.

[0019] Compared with the prior art, the present invention has the following beneficial effects: relative to conventional market maker services, the present high-frequency trading system can better realize functions such as enhancing market liquidity, improving price stability, preventing price manipulation, and market price discovery, and the present system takes computer vision algorithm as the core, and uses the market order queue data analysis and monitoring middleware as the "eyes" to build a decision-making system for monitoring abnormal transactions and order changes, which can realize technologies such as extremely fast market acquisition and analysis, market order pair recognition, active buy order volume recognition, large transaction order feature target tracking, abnormal transaction feature search, etc., and can carry multiple arbitrage strategies across markets, varieties, periods, etc. in parallel, and can realize high concurrency, low latency, and extremely fast transactions. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 is a structural block diagram of a high-frequency trading system based on artificial intelligence according to an embodiment of the present invention;

[0022] Figure 2 It is a structural block diagram of a global trading market information aggregation management system in a high-frequency trading system based on artificial intelligence according to an embodiment of the present invention;

[0023] Figure 3 It is a structural block diagram of a quantitative arbitrage trading hedging system in a high-frequency trading system based on artificial intelligence according to an embodiment of the present invention;

[0024] Figure 4 It is a structural block diagram of an order queue management system in a high-frequency trading system based on artificial intelligence according to an embodiment of the present invention;

[0025] Figure 5 is a structural block diagram of an arbitrage trading module in a high-frequency trading system based on artificial intelligence according to an embodiment of the present invention;

[0026] Figure 6 It is a structural block diagram of a core transaction queue algorithm controlling a matching engine pool module in an artificial intelligence-based high-frequency trading system according to an embodiment of the present invention.

[0027] Reference numerals:

[0028] 1. Global trading market quotation aggregation management system; 101. Extremely fast quotation acquisition module; 102. Monitoring middleware module; 103. Aggregation management module; 2. Quantitative arbitrage trading hedging system; 201. Transaction front-end module; 202. User module; 203. Network management module; 3. Risk control management system; 4. Quotation processing and analysis subsystem; 5. Order queue management system; 501. Order identification module; 502. Data analysis module; 6. Arbitrage trading module; 601. Order execution module; 602. Strategy module; 7. Core transaction queue algorithm control matching engine pool module; 701. Matching processing module; 702. Transaction record log module; 703. Memory data module; 8. Reverse hedging control engine module; 9. Account net value and quotation matching control module; 10. Transaction order management system; 11. Hedge trading market module. DETAILED DESCRIPTION

[0029] Below, the invention is further described in conjunction with the accompanying drawings and specific embodiments:

[0030] Example:

[0031] See also Figure 1-6 , according to an embodiment of the present invention, an artificial intelligence-based high-frequency trading system includes a global trading market aggregation management system 1, a quantitative arbitrage trading hedging system 2 and a risk control management system 3;

[0032] The output end of the global trading market market aggregation management system 1 is connected to a market processing and analysis subsystem 4, the output end of the market processing and analysis subsystem 4 is connected to an order queue management system 5, and the output end of the order queue management system 5 is connected to a plurality of arbitrage trading modules 6;

[0033] The output end of the quantitative arbitrage trading hedging system 2 is connected to a core trading queue algorithm control matching engine pool module 7, and the output end of the core trading queue algorithm control matching engine pool module 7 is connected to a reverse hedging control engine module 8;

[0034] The output end of the risk control management system 3 is connected to the account net value and market matching control module 9, the output end of the account net value and market matching control module 9 is connected to the transaction order management system 10, and the output end of the transaction order management system 10 is connected to several hedging transaction market modules 11;

[0035] The output end of the order queue management system 5 is connected to the core transaction queue algorithm control matching engine pool module 7, the output end of the account net value and market matching control module 9 is connected to the core transaction queue algorithm control matching engine pool module 7, and the output end of the reverse hedging control engine module 8 is connected to the transaction order management system 10.

[0036] In a further embodiment, the global trading market quotation aggregation management system 1 includes an extremely fast quotation acquisition module 101, a monitoring middleware module 102 and an aggregation management module 103. The extremely fast quotation acquisition module 101 is used to obtain quotation and data, and the quotation and data include historical data, real-time quotation, factor data, tick data and market data. The monitoring middleware module 102 is used to monitor market fluctuations, price spreads, cross-exchanges and background data. The aggregation management module 103 is used to aggregate market orders from multiple exchanges, and there are many market makers behind each exchange. Under the multiple market maker system, each pricing product has several market makers providing prices. The competition and constraints among market makers will make prices tend to the real value. This mechanism enables market makers to play the role of market prices.

[0037] In a further embodiment, the market processing and analysis subsystem 4 is capable of performing extremely fast market analysis, fully extracting market characteristics, and improving the algorithm's anti-noise capability based on algorithms and deep learning.

[0038] In a further embodiment, the order queue management system 5 includes an order identification module 501 and a data analysis module 502. The order identification module 501 is used to perform order queue identification, abnormal order identification, active buy order volume and sell order volume identification, and large transaction order feature identification. The data analysis module 502 can predict the queue order based on the market quotation of the machine learning algorithm. The order queue control and quotation determine the market arbitrage ability. The strategy is highly sensitive to the system response speed and delay. By buying and selling at high speed to capture the price fluctuations near the market bid and ask prices in a very short time, the strategy aims to accumulate the extremely small profits obtained from hundreds or thousands of transactions in a short period of time to achieve profitability.

[0039] In a further embodiment, the arbitrage trading module 6 includes an order execution module 601 and a strategy module 602. The order execution module 601 is used to execute pending orders, buried orders, placing orders and canceling orders. The strategy module 602 captures the price difference between two financial assets with exactly the same underlying assets to obtain profits through arbitrage strategies, high-frequency trading strategies and neutral strategies. Transaction aggregation, dark pool matching and high-frequency quantitative strategies are the three core weapons of this system. With the strong coordination of multiple tools, this system can capture the price difference opportunities of the entire market very well. With strong technical capabilities, ultra-low transaction fees and extremely low capital costs, it can ultimately achieve the goal of risk-free arbitrage in all scenarios and low capital costs.

[0040] In a further embodiment, the quantitative arbitrage trading hedging system 2 includes a trading front-end module 201, a user module 202 and a network management module 203. The trading front-end module 201 is used to realize the functions of monitoring transactions and customer link management. The user module 202 can generate and send terminal user-commissioned quotations and quantity orders to the trading platform. The network management module 203 is used to collect user orders and distribute the collected user orders to the core trading queue algorithm control matching engine pool module 7.

[0041] In a further embodiment, the core transaction queue algorithm controls the matching engine pool module 7 to include a matching processing module 701, a transaction record log module 702 and a memory data module 703. The matching processing module 701 is the core part of the trading system, which is used to receive orders and implement order matching according to business logic, generate transaction records at the same time, and then give users feedback on transaction results. It can realize buy and sell operations, short selling, closing positions, forced closing positions, increasing margin and other functions. An internal matching mechanism can be selected: internalizing customer orders, that is, using their own accounts or orders placed by other customers to directly trade with customers, thereby reducing the relevant costs of transactions on other exchange platforms. The transaction record log module 702 can record orders and transactions, and the memory data module 703 is used to store orders and transaction records during the transaction process to achieve data persistence.

[0042] In a further embodiment, the reverse hedging control engine module 8 can control its own risk exposure through hedging strategies, wherein the hedging strategy will control the risk exposure through the Greek letter values ​​of delta, gamma, vega, and thetahe rho. The most basic principle is to keep delta in the middle to avoid the impact of target price fluctuations on the system's positions. After the researcher completes the research on the hedging strategy, the program manager will implement the hedging strategy in the system. During the transaction, the system can automatically hedge according to the transaction situation, select other pricing contracts or target spot according to the hedging strategy model, and construct arbitrage combinations, pricing combinations, and pricing and spot combinations to achieve inventory risk control.

[0043] In a further embodiment, the risk control management system 3 can perform data monitoring, data analysis, exception handling and model iteration based on algorithms, and perform risk control, monitoring model risk, liquidity risk, operational risk, information asymmetry risk and survival risk before, during and after the event, so that the system can automatically monitor, manually assist, automatically warn and design circuit breakers, and use intelligent risk control to protect assets.

[0044] In a further embodiment, the account net value and market matching control module 9 is used to match the account net value and market conditions for trading, and the transaction order management system 10 module manages the transaction order information, wherein the transaction order information includes: transaction volume, transaction amount, average transaction price and transaction time, and the hedging trading market module 11 performs hedging when the strategy is triggered.

[0045] Through the above scheme of the present invention, compared with conventional market maker services, this high-frequency trading system can better realize the functions of enhancing market liquidity, improving price stability, preventing price manipulation, market price discovery, etc., and this system takes computer vision algorithm as the core, and uses the market quotation queue data analysis and monitoring middleware as the "eyes" to build a decision-making system for monitoring abnormal transactions and order changes, which can realize the rapid acquisition and analysis of market conditions, market quotation column recognition, active buy order volume recognition, large transaction order feature target tracking, abnormal transaction feature search and other technologies, and can carry multiple arbitrage strategies across markets, varieties, periods, etc. in parallel, and can realize high concurrency, low latency and extremely fast transactions.

[0046] In order to facilitate understanding of the above technical solutions of the present invention, the working principle or operation mode of the present invention in the actual process is described in detail below.

[0047] In actual application, this system takes machine vision algorithm as the core, involving technologies such as market order queue recognition, active buy order volume recognition, sell order volume recognition, large transaction order feature target tracking, abnormal transaction feature search, etc. It combines hedging arbitrage, high-frequency market making and dark pool matching technology, trying to aggregate the liquidity of order queue varieties, and ultimately in the trading market, continuously quote the best buying and selling prices (i.e. bilateral quotations) for specific varieties to the public traders, and accept the buying and selling requirements of public investors at the quoted prices, and trade with market counterparties with its own and customer entrusted funds, so as to achieve the purpose of risk-free arbitrage while completing the liquidity supply.

[0048] Compared with conventional market maker services, the aggregated trading market making system can better enhance market liquidity, improve price stability, prevent price manipulation, and market price discovery. The specific principles are as follows:

[0049] 1. With computer vision algorithm as the core, and using the market order queue data analysis and monitoring middleware as the "eyes", a risk control decision-making system dedicated to abnormal transactions and order change monitoring is built. Tens of thousands of "small eyes" enable the dragonfly's compound eyes to have accurate visual perception capabilities in market data changes. Based on machine vision algorithms, it can capture the high probability of market change trends, providing strong support for trading decisions and risk control;

[0050] 2. The time series anomaly detection algorithm based on visual saliency detection technology and deep learning has significantly improved the average precision, average recall and average f1 value compared with the baseline method. By applying the visual saliency technology in the field of computer vision, the saliency of each moment in the time series is effectively extracted, and the original time series is transformed into a form that is easier to analyze. In this case, the normal mode of the time series can be better captured by the neural network, so that anomalies can be detected more effectively;

[0051] 3. This system combines hedging arbitrage, high-frequency market making and dark pool matching technologies, attempting to aggregate the liquidity of order queue varieties, and ultimately continuously quote the best buying and selling prices (i.e., bilateral quotations) for specific varieties to public traders in the trading market, and accept the buying and selling requirements of public investors at the quoted prices, and trade with market counterparties with its own and customer entrusted funds, thereby achieving the purpose of risk-free arbitrage while completing the liquidity supply.

[0052] This system takes computer vision algorithm as its core, and uses the market order queue data analysis and monitoring middleware as its "eyes" to build a decision-making system for monitoring abnormal transactions and order changes.

[0053] It can realize: extremely fast market acquisition and analysis, market quotation and order pair recognition, active buy order volume recognition, large transaction order feature target tracking, abnormal transaction feature search and other technologies, intelligent strategy, intelligent risk control, high-frequency trading, statistical arbitrage across periods, products, markets, and market making.

[0054] Parallel: Equipped with multiple arbitrage strategies across markets, products, and periods, the traditional arbitrage strategy is combined with the computer C++ program underlying code to develop memory management, and the advantages of high frequency and low latency in high-precision time synchronization. The strategy arbitrage spread is compressed into a space that conventional trading teams cannot compete in, maintaining the company's leading technical capabilities in the industry.

[0055] High concurrency, low latency and ultra-fast trading: By optimizing computer network hardware, single-chip algorithm writing and other systems, we have achieved a communication link from market acquisition and analysis → strategy triggering → order engine execution → transaction execution through the network card. The hardware of the entire execution process has made comprehensive execution speed optimization from memory, GPU, network card, underlying control calls and other links, so that the strategy execution rate remains among the leaders in the industry.

[0056] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A high-frequency trading system based on artificial intelligence, characterized in that: It includes a global trading market aggregation management system (1), a quantitative arbitrage trading hedging system (2) and a risk control management system (3); The output end of the global trading market market aggregation management system (1) is connected to a market processing and analysis subsystem (4), the output end of the market processing and analysis subsystem (4) is connected to an order queue management system (5), and the output end of the order queue management system (5) is connected to a plurality of arbitrage trading modules (6); The output end of the quantitative arbitrage trading hedging system (2) is connected to a core trading queue algorithm control matching engine pool module (7), and the output end of the core trading queue algorithm control matching engine pool module (7) is connected to a reverse hedging control engine module (8); The output end of the risk control management system (3) is connected to an account net value and market matching control module (9), the output end of the account net value and market matching control module (9) is connected to a transaction order management system (10), and the output end of the transaction order management system (10) is connected to a plurality of hedging transaction market modules (11); The output end of the order queue management system (5) is connected to the core transaction queue algorithm control matching engine pool module (7), the output end of the account net value and market matching control module (9) is connected to the core transaction queue algorithm control matching engine pool module (7), and the output end of the reverse hedging control engine module (8) is connected to the transaction order management system (10); The global trading market information aggregation management system (1) comprises an extremely fast information acquisition module (101), a monitoring middleware module (102) and an aggregation management module (103), wherein the extremely fast information acquisition module (101) is used to acquire information and data, wherein the information and data include historical data, real-time information, factor data, tick data and market data, the monitoring middleware module (102) is used to monitor market fluctuations, market spreads, cross-exchange monitoring and background data monitoring, and the aggregation management module (103) is used to aggregate market orders from multiple exchanges; The order queue management system (5) comprises an order identification module (501) and a data analysis module (502). The order identification module (501) is used to identify order queues, abnormal orders, active buy order quantities and sell order quantities, and large transaction order feature identification. The data analysis module (502) can predict queue orders based on the market quotation of the machine learning algorithm. The order queue control and quotation determine the market arbitrage ability. The strategy is highly sensitive to the system response speed and delay. The price fluctuations near the market bid and ask prices in a very short time are captured by high-speed buying and selling. The arbitrage trading module (6) comprises an order execution module (601) and a strategy module (602), wherein the order execution module (601) is used to execute pending orders, deposit orders, place orders and cancel orders, and the strategy module (602) captures the price difference between two financial assets with exactly the same underlying assets through arbitrage strategies, high-frequency trading strategies and neutral strategies to obtain profits; The quantitative arbitrage trading hedging system (2) comprises a trading front-end module (201), a user module (202) and a network management module (203), wherein the trading front-end module (201) is used to implement monitoring trading and customer link management functions, the user module (202) can generate and send terminal user commissioned quotations and quantity orders to the trading platform, and the network management module (203) is used to collect user orders and distribute the collected user orders to the core trading queue algorithm control matching engine pool module (7); The core transaction queue algorithm controls the matching engine pool module (7) and includes a matching processing module (701), a transaction record log module (702) and a memory data module (703). The matching processing module (701) is the core part of the transaction system and is used to receive orders and match orders according to business logic, generate transaction records, and then provide transaction result feedback to users. The transaction record log module (702) can record orders and transactions. The memory data module (703) is used to store orders and transaction records during the transaction process to achieve data persistence. The reverse hedging control engine module (8) can control its own risk exposure through hedging strategies, wherein the hedging strategy will control the risk exposure through the Greek letter values ​​of delta, gamma, vega, and thetahe rho. The most basic principle is to keep delta in the middle to avoid the impact of the underlying price fluctuation on the system's positions. After the researcher completes the research on the hedging strategy, the program manager will implement the hedging strategy in the system. During the transaction process, the system can automatically hedge according to the transaction situation, select other pricing contracts or underlying spot according to the hedging strategy model, and construct arbitrage combinations, pricing combinations, and pricing and spot combinations to achieve inventory risk control.

2. The high-frequency trading system based on artificial intelligence according to claim 1, characterized in that: The market processing and analysis subsystem (4) is capable of performing extremely fast market analysis, fully extracting market features, and improving the algorithm's anti-noise capability based on algorithms and deep learning.

3. The high-frequency trading system based on artificial intelligence according to claim 1, characterized in that: The risk control management system (3) can perform data monitoring, data analysis, exception handling and model iteration based on algorithms, and perform risk control, monitoring model risk, liquidity risk, operational risk, information asymmetry risk and survival risk before, during and after the event.

4. The high-frequency trading system based on artificial intelligence according to claim 1, characterized in that: The account net value and market matching control module (9) is used to match the account net value and market conditions for trading, the transaction order management system (10) module manages transaction order information, and the hedging transaction market module (11) performs hedging when a strategy is triggered.

Images