CN108229563A - A kind of treasury bond futures actively do city's system - Google Patents
A kind of treasury bond futures actively do city's system Download PDFInfo
- Publication number
- CN108229563A CN108229563A CN201810007501.0A CN201810007501A CN108229563A CN 108229563 A CN108229563 A CN 108229563A CN 201810007501 A CN201810007501 A CN 201810007501A CN 108229563 A CN108229563 A CN 108229563A
- Authority
- CN
- China
- Prior art keywords
- module
- data
- exchange
- model
- transaction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/25—Integrating or interfacing systems involving database management systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/54—Interprogram communication
- G06F9/546—Message passing systems or structures, e.g. queues
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/04—Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/06—Asset management; Financial planning or analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/54—Indexing scheme relating to G06F9/54
- G06F2209/547—Messaging middleware
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Marketing (AREA)
- Human Resources & Organizations (AREA)
- General Engineering & Computer Science (AREA)
- General Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Operations Research (AREA)
- Technology Law (AREA)
- Entrepreneurship & Innovation (AREA)
- Databases & Information Systems (AREA)
- Game Theory and Decision Science (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The invention discloses a kind of treasury bond futures actively to do city's system, city's system of doing includes exchange to connection module, server cluster control module, message-oriented middleware module and database storage module, wherein, the exchange obtains market data and transaction data from exchange's gateway to connection module and establishes buffer queue；The server cluster is converted into the algorithm model for meeting strategy, output model result from the exchange to the buffer queue extraction related data of connection module；The message-oriented middleware module will be sent to database storage module positioned at bottom after various types of messages data summarization that server cluster control module generates；The database storage module stores the message data.City's system disclosed in this invention of doing steady real-time can be handled multidimensional marketplace relationship, can be carried out dynamic self-teaching and amendment with the variation in market, then provide and do city's information using hardware type distributed architecture.
Description
Technical field
The invention belongs to financial instruments and futures field, and in particular to a kind of treasury bond futures actively do city's system.
Background technology
The treasury bond futures Interest rate futures contract important as one kind, to the price discovery of market rate, interest rate risk shifts,
Important function is suffered from terms of improving efficiency in the use of funds and diversified Asset Allocation.
Under the background for especially carrying out interest rate market-oriented reform in current China, the interest rate of treasury bond futures how is given full play to
The effect of price discovery, interest rate risk transfer etc., will directly influence the effect of interest rate market-oriented reform.It is and better
Play these functions of treasury bond futures, it is necessary to the preferable market liquidity, however treasury bond futures market flowing domestic at present
Property also there are it is apparent the defects of, be mainly shown as that the treasury bond futures mechanism participation of the current country is inadequate, also deposited in terms of mobility
In deficiency, it is then a kind of common method for solving flowability problem to do city.
At present, on the one hand, it is existing based on treasury bond futures to do city's system seldom；On the other hand, tradition does city's method, by
Be limited to data storage, data-handling capacity and algorithm design aspect, there are it is larger the problem of, for example historical information excavation do not fill
Point, real-time is inadequate etc.；In addition, part system is using manual transaction Yuan Zuo cities, but so it is difficult to ensure that stablize can the moment
By performing.
Invention content
Drawbacks described above based on the prior art, the purpose of the present invention is to provide a kind of treasury bond futures actively to do city's system,
By using the distributed hardware framework in forward position, the hot standby of hardware ensure that, improve the hardware reliability of system, use simultaneously
Multithreading, multi-course concurrency technology accelerate with reference to GPU, improve rapid computations ability of the program to mass data, same with this
When, improvement has also been made in we in terms of city's algorithm is done, can be original by using new machine learning and deep learning method
Mass historical data preferably use so that system can preferably provide mobility, while control itself risk.
A kind of treasury bond futures of the present invention actively do city's system, including exchange to connection module, server set group control mould
Block, message-oriented middleware module and database storage module, wherein,
The exchange connects the server cluster control module and exchange's gateway, the exchange to connection module
Market data and transaction data are obtained from exchange's gateway to connection module and establish buffer queue；
The server cluster control module connects the exchange and forwards mould to connection module and message-oriented middleware respectively
Block, the server cluster are converted into the calculation for meeting strategy from the exchange to the buffer queue extraction related data of connection module
Method model, output model result；
The message-oriented middleware module connects the server cluster control module and database storage module, institute respectively
Stating message-oriented middleware module will be sent to after various types of messages data summarization that server cluster control module generates positioned at bottom
Database storage module；
The database storage module connects the message-oriented middleware forwarding module, stores the message data.
Further, in the preferred scheme of said one, the exchange includes market module and transaction to connection module
Module, the market module and transaction modules are respectively provided with interface corresponding with exchange gateway, and the market module is used for from institute
It states exchange's gateway and obtains market data, and establish market buffer queue；The transaction modules are used for from exchange's gateway
It obtains transaction data and the transaction data is put into transaction buffer queue, meanwhile, also receive the server set group control mould
The buying signals that block generates, and the data structure for meeting exchange's gateway requirement is built according to the buying signals, it calls later
The API of exchange's gateway sends transaction request.
Further, in the preferred scheme of said one, the server cluster control module includes server cluster
And policing algorithm model module, the server cluster extracts market data from market buffer queue, and market data are carried out
Format conversion is trained new input excellent into the input form for meeting policing algorithm model, the policing algorithm model module
After changing generation optimization, new model result is exported.
Further, in the preferred scheme of said one, the market module further includes data filtering module, for pair
The advanced row data screening of market data of reception and filtering, and the data after filtering are sent to the server set group control mould
Block.
Further, in the preferred scheme of said one, the exchange further includes connection module message logging classification
Module generates message logging queue to the market data and transaction data of generation respectively, and passes through special message logging line
Message logging is sent to database storage module and stored by journey.
Further, in the preferred scheme of said one, the policing algorithm model module includes：Historical data stream list
Member, model training cluster unit, model result storage unit and transactional services cluster, wherein,
The historical data stream unit connects the model training cluster unit and model result storage unit, institute respectively
It states historical data stream unit and preserves the data after over cleaning, and historical data is sent to model training cluster unit, it is described
Historical data stream includes at least historical quotes factor data, news public sentiment factor data and macro-data；
The model training cluster unit connects the historical data stream unit, model result storage unit and friendship respectively
Easy service cluster, the model training cluster unit is from the historical data stream cell call historical data, while described in receiving
After transactional services cluster send back the transaction results come, historical data and transaction results are loaded into model and constantly trained,
And the model result for completing training is sent to model result storage unit；
The model result storage unit, for preserving trained model result, while the history number that model is needed
It carries out being initially loaded caching according to index；
The transactional services cluster after the model result storage unit carries out model result update, triggers the friendship
Easy service cluster calls newest model, and be traded meter with reference to newest transaction data from the model result storage unit
It calculates.
Further, in the preferred scheme of said one, the model training cluster unit is by SVM training patterns, GLM
Training pattern and RNN training patterns composition, wherein, the SVM training patterns in the model training cluster unit are right in advance
Historical data and transaction results carry out classification based training, then recall the GLM training patterns and classification results are carried out with quantitative tune
Whole, the RNN training patterns group then carries out parallel training with aforementioned SVM training patterns and GLM training patterns simultaneously.
Further, in the preferred scheme of said one, it is described do city's system and further include connect the exchange respectively
To the configuration module of connection module and service device clustered control module, the configuration module is used for exchange to connection module and service
Device clustered control mould data configuration in the block.
Further, in the preferred scheme of said one, the server cluster control module also caches team toward message
Related data is written according to form set in advance in row, including entrusted information, deal message and error message, and is sent to message
Middleware module, the message-oriented middleware module are sent to database storage module after the various information of reception is integrated.
Compared with prior art, a kind of treasury bond futures disclosed in this invention actively do city's system, and one side is overturned
Traditional does city's method, and passive-type is done city becomes actively to choose whether to do city, on the other hand, by exchange to connection module,
The hardware distribution framework of server cluster control module, message-oriented middleware module and database storage module, can be real-time
Steady processing multidimensional marketplace relationship, secondly, sets policing algorithm model in server cluster control module, can be with
The variation in market carries out dynamic self-teaching and amendment.Simultaneously using machine learning models such as SVM, the RNN established to magnanimity
High-frequency data handled real-time, then provide and do city's information, while itself risk is both controlled, carried for market
For certain mobility.
Description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, to embodiment or will show below
There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair
Some bright embodiments, for those of ordinary skill in the art, without creative efforts, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is the hardware architecture diagram that a kind of treasury bond futures actively do city's system in one embodiment of the invention.
Fig. 2 is the structure principle chart of server cluster control module in one embodiment of the invention.
Fig. 3 is the operation logic figure about RNNs models in one embodiment of the invention.
Specific embodiment
For those skilled in the art is made to more fully understand technical scheme of the present invention, below in conjunction with the accompanying drawings and specific embodiment party
Formula is described in further detail the present invention.However, it should be understood that these descriptions are merely illustrative, and it is not intended to limit this hair
Bright range.In addition, in the following description, the description to known features and technology is omitted, to avoid this is unnecessarily obscured
The concept of invention.
With reference to shown in Fig. 1, a kind of treasury bond futures disclosed in the embodiment of the present invention actively do city's system, including exchange pair
Connection module 1, server cluster control module 2, message-oriented middleware module 3 and database storage module 4, wherein,
The exchange connects the server cluster control module 2 and exchange's gateway, the transaction to connection module 1
Institute obtains market data and transaction data from exchange's gateway to connection module 1 and establishes buffer queue；
The server cluster control module 2 connects the exchange to connection module 1 and message-oriented middleware module 3 respectively,
The server cluster control module 2 is converted into meeting from the exchange to the buffer queue extraction related data of connection module 1
The algorithm model of strategy, output model result；
The message-oriented middleware module 3 connects the server cluster control module 2 and database storage module 4 respectively,
The message-oriented middleware module 3 will be sent to after various types of messages data summarization that server cluster control module 2 generates the bottom of positioned at
The database storage module 4 of layer；
The database storage module 4, connects the message-oriented middleware forwarding module, and storage is sent by message-oriented middleware
All kinds of message datas to come over.
Specifically, in one embodiment of the invention, the exchange includes market module and transaction to connection module 1
Module, the two are respectively provided with the interface docked with exchange gateway, i.e. market interface and transaction interface.The market module is led to
The market data that market interface exchange gateway is sent are crossed, and establish market buffer queue, so as to ensure the market of system
Thread will not block, wherein, the market data mainly with itself according to tick market process by the tick market including exchange
Minute K line number evidence.Preferably, the market module is also interior is equipped with data filtering module, first for the market data to reception
Data screening and filtering are carried out, for the nontransaction time, price is filtered beyond the data of price limits range etc, and will
Data after filtering are sent to the server cluster control module 2, for aspect of model construction and screening of next step etc..Institute
Transaction modules are stated, the transaction data sent by transaction interface reception from exchange's gateway, the transaction data includes report
List, removes the information such as list, mistake at exchange hand.The transaction data of reception is put into transaction buffer queue by the transaction modules, to ensure
The transaction thread of system will not be blocked, meanwhile, the transaction modules are also responsible for receiving the friendship that service cluster control module 2 generates
Easy signal, buying signals herein are primarily referred to as the strategy operation module inside service cluster control module 2 received
Whether mono signal under the dealing of generation including lower single direction, lower single quantity, the commission stand-by period, needs to remove the inferior information of substance,
And meet the data structure of exchange's gateway requirement according to buying signals structure, the API of exchange's gateway is called later, is sent and is handed over
Easily request.
The exchange further includes message logging sort module to connection module 1, respectively to the market number of generation
Message logging queue is generated according to transaction data, and passes through special, corresponding message logging thread by message logging (comprising row
Feelings data and transaction data) be sent to database storage module 4 carry out storage file.
The server cluster control module 2, including server cluster and policing algorithm model module, the server set
For group again comprising quotation service device cluster and trading server cluster, the policing algorithm model module is substantially also to pass through clothes
Business device cluster realizes the iteration update of algorithm, and the server cluster extracts the market buffer queue of connection module 1 from exchange
Market data, and market data are subjected to format conversion into the input form for meeting policing algorithm model, since initial data is
Structure form, algorithm need data be the form of vector or map, therefore need carry out format conversion, market data and
Transaction data is all the data that policing algorithm model needs, and market data are input variables, and transaction results are predictive variables, model
To be studied by analyzing the relationship of market data and transaction results, the policing algorithm model module to new input into
After row training optimization, new model result is exported, new model result will cover old model, then supply policy object tune again
With.The server cluster control module 2 also has with exchange to the communication port of connection module 1 and message-oriented middleware module 3,
On the one hand, the server cluster control module 2 sends out the buying signals after the training optimization of policing algorithm model module again
Exchange is given to connection module 1, on the other hand, the server cluster control module 2 is also toward message buffer queue according to prior
The form write-in related data of setting, including entrusted information, deal message and error message, and is sent to message-oriented middleware module
3, the message-oriented middleware module 3 is sent to database storage module 4 after the various information of reception is integrated.
The message-oriented middleware module 3, mainly by various types of messages, including including entrusted information, deal message and mistake
Information is uniformly aggregated to form a unified module and is then sent to bottom data library storage.
The database storage module 4 receives the Various types of data that message-oriented middleware module 3 is sent, and can be used for storage knot
Structure data and unstructured data, such as media event data.
Wherein, each exchange is to connection module 1, server cluster control module 2, in message-oriented middleware module 3 also
Cache module is built-in with, guarantee system will not excessively block.
In one embodiment of the invention, it is described actively do city's system and further include connect exchange docking mould respectively
The configuration module of block 1 and server clustered control module 2, the configuration module are used for exchange to connection module 1 and server
Data configuration in clustered control module 2.The configuration module is also embedded in inside modules.Data configuration is included to each
The parameter that a modular system initialization needs is configured, such as account/password, corresponding transaction gateway etc..
With reference to shown in Fig. 2, the policing algorithm model module mainly completes the iteration update of data, by using new machine
Device learns and deep learning method, original mass historical data can be preferably used so that system can be more preferable
Offer mobility, while control itself risk.It includes：Historical data stream unit, model training cluster unit, model knot
Fruit storage unit and transactional services cluster, wherein,
The historical data stream unit connects the model training cluster unit and model result storage unit, institute respectively
It states historical data stream unit and preserves the data after over cleaning, cleaning herein includes removing repeated data, deletes the nontransaction time
Market data, delete the data outside price limits valency, delete average price and be more than data etc. of price limits valency, and historical data is sent
Model training cluster unit is given, the historical data stream is poor including at least historical quotes factor data, such as dealing commission amount, tires out
Accumulated amount difference etc.；News public sentiment factor data, such as kind contract attention rate, contract are held position ranking etc.；Macro-data, such as bank
Between make a short-term loan on interest rate, national debt is against repo rate etc.；
The model training cluster unit connects the historical data stream unit, model result storage unit and friendship respectively
Easy service cluster, the model training cluster unit is on the one hand from the historical data stream cell call historical data, and simultaneously
After receiving the transaction results that the transactional services cluster is sent back, by historical data and this two block number of transaction results according to loading mould
Type is constantly trained, and the model result that training is completed is sent to model result storage unit.
The model result storage unit, for preserving trained model result, while the history number that model is needed
It carries out being initially loaded caching according to index, avoids frequent data call that server is caused to bear.Wherein, the model storage unit
Mainly using internal storage data library storage, aforementioned system database memory module is mainly using data in magnetic disk library storage.
The transactional services cluster has multiple servers, primarily to ensure the stable and high effective operation of transaction modules,
And there is disaster-tolerant backup, after the model result storage unit carries out model result update, trigger the transaction clothes
Business cluster calls newest model (after model often updates once, can trigger primary new tune from the model result storage unit
With), and it is traded calculating with reference to newest transaction data.The transactional services cluster is provided with special network channel, adopts
Transaction results are real-time transmitted to model training cluster unit with private network communication.
In one embodiment of the invention, the model training cluster unit is by SVM training patterns, GLM training patterns
And RNN training patterns composition, wherein, the SVM training patterns in the model training cluster unit are in advance to historical data
Classification based training is carried out with transaction results, the GLM training patterns is then recalled and quantitative adjusting is carried out to classification results, and it is described
RNN training patterns group then carries out parallel training with aforementioned SVM training patterns and GLM training patterns simultaneously, it is mainly responsible for friendship
The training optimization of easy execution module, including lower single mode, stand-by period etc..Pass through SVM training patterns, GLM training patterns and RNN
Training pattern carries out continuous training and iteration update.For the present invention by blending algorithm, system is basic to industry to different models
The accuracy of the prediction result in face forms dynamic and judges, achievees the purpose that dynamic optimization model, allows entire prediction model result
Accuracy more stablize, and whole system from data collection to generate prediction result all completed by computer, greatly
The time and efforts of analyst is liberated.
Introduce the calculation in the embodiment of the present invention about SVM training patterns, GLM training patterns and RNN training patterns below
Method principle.
Wherein, SVM (support vector machines, Support Vector Machine) model, support vector machine method are to establish
In the VC dimensions theory and Structural risk minization basis of Statistical Learning Theory, according to limited sample information in model
Complexity (i.e. to the study precision of specific training sample, Accuracy) and learning ability (identify arbitrary sample without error
Ability) between seek optimal compromise, to obtain best Generalization Ability (or generalization ability).
SVM wishes the hyperplane by following form to be divided to sample：
G (x)=wx+b
It is 0 that we, which can take threshold value, and in this way when needing to differentiate there are one sample xi, we just see the value of g (xi).
If g (xi)>0, classification C1 is just determined as, if g (xi)<0, then it is determined as classification C2.
The solution of SVM, which is equivalent to, solves following optimization problems：
GLM (generalized linear model, generalized linear model) model is directly pushing away for general linear model
Extensively, it makes the population mean of dependent variable depend on linear prediction by a non-linear contiguous function (link function)
Value, while also response probability is allowed to be distributed as any a member in exponential family of distributions.Many widely applied statistical models belong to
Mould is returned in generalized linear model, such as logistic regression models, Probit regression models, Poisson regression models, negative binomial
Type etc..
The probability density (probability function) of exponential family of distributions is represented by：
Wherein, θ and φ is two parameters, and θ is known as natural parameter, and φ is discrete parameter；A, b, c are function.
E (y)=μ=b'(θ)
Var (y)=φ b " (θ)
The purpose of RNN is using carrying out processing sequence data.It is from input layer to hidden layer in traditional neural network model
Output layer is arrived again, is connected entirely between layers, and the node between every layer is connectionless.But this common nerve net
Network is for many problems but helpless.For example, you will predict that next word of sentence is, generally require and use front
Word because front and rear word is not independent in a sentence.Why RNNs is known as cycle neural network, i.e. a sequence
The output of output and front before broomrape is also related.The specific form of expression can be remembered and answered to the information of front for network
For in the calculating that currently exports, i.e., the node between hidden layer to be no longer connectionless but has connection, and hidden layer is defeated
Enter the output that the not only output including input layer further includes last moment hidden layer.Theoretically, RNNs can be to any length
Sequence data is handled.
Fig. 3 is the structure chart of a typical RNNs model in the embodiment of the present invention.
In Fig. 3, xt represents t, t=1,2, the input of 3... steps (step)；St is the state that the t of hidden layer is walked, it
It is the mnemon of network；St is calculated according to the output and the state of previous step hidden layer of current input layer.St=f (Uxt
+ Wst-1), wherein f is usually nonlinear activation primitive, and such as tanh or ReLU, when calculating s0, i.e., first word is hidden
Hide layer state, need to use s-1, but itself and be not present, be generally set to 0 vector in the implementation；Ot is the output of t steps, such as
The vector expression of next word, ot=softmax (Vst)
The blending algorithm that the present invention is formed by SVM training patterns, GLM training patterns and RNN training patterns, to different moulds
Type forms the accuracy of the prediction result of industry basic side dynamic and judges, achievees the purpose that dynamic optimization model, allows entire
The accuracy of prediction model result is more stablized, and while both having controlled itself risk, certain mobility is provided for market.
City's system of the present invention of doing steady processing multidimensional marketplace can be closed real-time using hardware distribution framework
Secondly system, sets policing algorithm model in server cluster control module, can be with the variation in market, into Mobile state
Self-teaching and amendment, then provide and do city's information.
Finally it should be noted that：The above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
The present invention is described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that：It still may be used
To modify to the technical solution recorded in foregoing embodiments or carry out equivalent replacement to which part technical characteristic；
And these modification or replace, various embodiments of the present invention technical solution that it does not separate the essence of the corresponding technical solution spirit and
Range.
Claims (9)
1. a kind of treasury bond futures actively do city's system, which is characterized in that city's system of doing includes exchange to connection module, service
Device clustered control module, message-oriented middleware module and database storage module, wherein,
The exchange connects the server cluster control module and exchange's gateway, exchange's docking to connection module
Module obtains market data and transaction data from exchange's gateway and establishes buffer queue；
The server cluster control module connects the exchange to connection module and message-oriented middleware forwarding module, institute respectively
It states server cluster and the algorithm mould for meeting strategy is converted into the buffer queue extraction related data of connection module from the exchange
Type, output model result；
The message-oriented middleware module connects the server cluster control module and database storage module respectively, described to disappear
Breath middleware module will be sent to data positioned at bottom after various types of messages data summarization that server cluster control module generates
Library storage module；
The database storage module connects the message-oriented middleware forwarding module, stores the message data.
2. treasury bond futures according to claim 1 actively do city's system, which is characterized in that the exchange is to connection module packet
Include market module and transaction modules, the market module and transaction modules are respectively provided with interface corresponding with exchange gateway, described
Market module is used to obtain market data from exchange's gateway, and establish market buffer queue；The transaction modules are used for
Transaction data is obtained from exchange's gateway and the transaction data is put into transaction buffer queue, meanwhile, it also receives described
The buying signals that server cluster control module generates, and the number for meeting exchange's gateway requirement is built according to the buying signals
According to structure, the API of exchange's gateway is called later, sends transaction request.
3. treasury bond futures according to claim 2 actively do city's system, which is characterized in that the server set group control mould
Block includes server cluster and policing algorithm model module, and the server cluster extracts market data from market buffer queue,
And market data are subjected to format conversion into the input form for meeting policing algorithm model, the policing algorithm model module is to new
Input be trained optimization generation optimization after, export new model result.
4. treasury bond futures according to claim 2 actively do city's system, which is characterized in that the market module further includes number
According to filtering module, for the advanced row data screening of market data to reception and filtering, and the data after filtering are sent to institute
State server cluster control module.
5. treasury bond futures according to claim 2 actively do city's system, which is characterized in that the exchange to connection module also
Including message logging sort module, message logging queue is generated to the market data and transaction data of generation respectively, and pass through
Message logging is sent to database storage module and stored by special message logging thread.
6. treasury bond futures according to claim 3 actively do city's system, which is characterized in that the policing algorithm model module
Including：Historical data stream unit, model training cluster unit, model result storage unit and transactional services cluster, wherein,
The historical data stream unit connects the model training cluster unit and model result storage unit respectively, described to go through
History data stream element preserves the data after over cleaning, and historical data is sent to model training cluster unit, the history
Data flow includes at least historical quotes factor data, news public sentiment factor data and macro-data；
The model training cluster unit connects the historical data stream unit, model result storage unit and transaction clothes respectively
Business cluster, the model training cluster unit receive the transaction from the historical data stream cell call historical data
After service cluster sends back the transaction results come, historical data and transaction results are loaded into model and constantly trained, and will
The model result that training is completed is sent to model result storage unit；
The model result storage unit for preserving trained model result, while the historical data that model needs is referred to
Mark carries out being initially loaded caching；
The transactional services cluster after the model result storage unit carries out model result update, triggers the transaction clothes
Cluster be engaged in from the newest model of model result storage unit calling, and calculating is traded with reference to newest transaction data.
7. treasury bond futures according to claim 6 actively do city's system, which is characterized in that the model training cluster unit
It is made of SVM training patterns, GLM training patterns and RNN training patterns, wherein, it is described in the model training cluster unit
SVM training patterns carry out classification based training to historical data and transaction results in advance, then recall the GLM training patterns to dividing
Class result carry out quantitative adjusting, the RNN training patterns group then simultaneously with aforementioned SVM training patterns and GLM training patterns into
Row parallel training.
8. treasury bond futures according to claim 6 actively do city's system, which is characterized in that city's system of doing further includes point
Configuration module of the exchange to connection module and service device clustered control module is not connected, and the configuration module is used for transaction
Institute is to connection module and service device clustered control mould data configuration in the block.
9. treasury bond futures according to claim 2 actively do city's system, which is characterized in that the server set group control mould
Also related data is written according to form set in advance toward message buffer queue in block, including entrusted information, deal message and mistake
Information, and message-oriented middleware module is sent to, the message-oriented middleware module is sent to number after the various information of reception is integrated
According to library storage module.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810007501.0A CN108229563A (en) | 2018-01-04 | 2018-01-04 | A kind of treasury bond futures actively do city's system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810007501.0A CN108229563A (en) | 2018-01-04 | 2018-01-04 | A kind of treasury bond futures actively do city's system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108229563A true CN108229563A (en) | 2018-06-29 |
Family
ID=62645235
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810007501.0A Pending CN108229563A (en) | 2018-01-04 | 2018-01-04 | A kind of treasury bond futures actively do city's system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108229563A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109447806A (en) * | 2018-11-01 | 2019-03-08 | 中国银行股份有限公司 | Front-end control method and device, storage medium and electronic equipment |
| CN109816529A (en) * | 2019-01-31 | 2019-05-28 | 黄涛 | A kind of digital cash unifies method of commerce |
| CN110619580A (en) * | 2019-09-06 | 2019-12-27 | 北京神州同道智能信息技术有限公司 | Whole-market multi-variety fund financing management system based on intelligent data processing platform |
| CN111161060A (en) * | 2019-11-28 | 2020-05-15 | 北京睿智融科控股股份有限公司 | Comprehensive platform for investment and research transaction |
| CN112669148A (en) * | 2020-12-22 | 2021-04-16 | 深圳市富途网络科技有限公司 | Order processing method and device |
| CN114565465A (en) * | 2022-03-08 | 2022-05-31 | 中国工商银行股份有限公司 | Automated transaction method, apparatus, device, storage medium and program product |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1967579A (en) * | 2006-02-13 | 2007-05-23 | 湖南大学 | Aid excavating analysis system of financial risk management |
| US7536332B2 (en) * | 2001-02-02 | 2009-05-19 | Rhee Thomas A | Real life implementation of modern portfolio theory (MPT) for financial planning and portfolio management |
| US7640206B1 (en) * | 2003-12-12 | 2009-12-29 | Trading Technologies International, Inc. | System and method for event-based trading |
| CN102663649A (en) * | 2012-05-18 | 2012-09-12 | 苏州工业园区凌志软件有限公司 | Financial derivative transaction system |
-
2018
- 2018-01-04 CN CN201810007501.0A patent/CN108229563A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7536332B2 (en) * | 2001-02-02 | 2009-05-19 | Rhee Thomas A | Real life implementation of modern portfolio theory (MPT) for financial planning and portfolio management |
| US7640206B1 (en) * | 2003-12-12 | 2009-12-29 | Trading Technologies International, Inc. | System and method for event-based trading |
| CN1967579A (en) * | 2006-02-13 | 2007-05-23 | 湖南大学 | Aid excavating analysis system of financial risk management |
| CN102663649A (en) * | 2012-05-18 | 2012-09-12 | 苏州工业园区凌志软件有限公司 | Financial derivative transaction system |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109447806A (en) * | 2018-11-01 | 2019-03-08 | 中国银行股份有限公司 | Front-end control method and device, storage medium and electronic equipment |
| CN109816529A (en) * | 2019-01-31 | 2019-05-28 | 黄涛 | A kind of digital cash unifies method of commerce |
| CN110619580A (en) * | 2019-09-06 | 2019-12-27 | 北京神州同道智能信息技术有限公司 | Whole-market multi-variety fund financing management system based on intelligent data processing platform |
| CN110619580B (en) * | 2019-09-06 | 2023-11-24 | 北京神州同道智能信息技术有限公司 | Whole market multi-variety gold financing system based on intelligent data processing platform |
| CN111161060A (en) * | 2019-11-28 | 2020-05-15 | 北京睿智融科控股股份有限公司 | Comprehensive platform for investment and research transaction |
| CN112669148A (en) * | 2020-12-22 | 2021-04-16 | 深圳市富途网络科技有限公司 | Order processing method and device |
| CN114565465A (en) * | 2022-03-08 | 2022-05-31 | 中国工商银行股份有限公司 | Automated transaction method, apparatus, device, storage medium and program product |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108229563A (en) | A kind of treasury bond futures actively do city's system | |
| Yan et al. | Research on financial assets transaction prediction model based on LSTM neural network | |
| Li et al. | A review of applications in federated learning | |
| US9870630B2 (en) | Methods and systems to generate graphical representations of relationships between persons based on transactions | |
| Soramäki et al. | The topology of interbank payment flows | |
| CN107705149A (en) | Data method for real-time monitoring, device, terminal device and storage medium | |
| CN110378786B (en) | Model training method, default transmission risk identification method, device and storage medium | |
| CN108615096A (en) | Server, the processing method of Financial Time Series and storage medium | |
| Lv et al. | [Retracted] An Economic Forecasting Method Based on the LightGBM‐Optimized LSTM and Time‐Series Model | |
| CN108600320A (en) | A kind of data cache method, apparatus and system | |
| CN109684087A (en) | Operation method, device and Related product | |
| CN112215616B (en) | Method and system for automatically identifying abnormal fund transaction based on network | |
| Vu et al. | Distributed adaptive model rules for mining big data streams | |
| CN109670927A (en) | The method of adjustment and its device of credit line, equipment, storage medium | |
| CN109710668A (en) | A kind of multi-source heterogeneous data access middleware construction method | |
| Ho et al. | A network analysis of the cryptocurrency market | |
| Sha | Time Series Stock Price Forecasting Based on Genetic Algorithm (GA)-Long Short-Term Memory Network (LSTM) Optimization | |
| Zhang et al. | A novel Bitcoin and Gold prices prediction method using an LSTM‐P neural network model | |
| Liu et al. | Learning-based adaptive data placement for low latency in data center networks | |
| Lian et al. | A blockchain prediction model on time, value, and purchase based on Markov chain and queuing theory in Stock trade | |
| Xie | Study on enterprise financial risk prevention and early warning system based on blockchain technology | |
| CN113220447A (en) | Financial wind control system and method based on edge calculation | |
| Li et al. | Strengthen credit scoring system of small and micro businesses with soft information: Analysis and comparison based on neural network models | |
| Li et al. | High-tech industries and competitive advantage in emerging markets: a study of foreign telecommunications equipment firms in China | |
| CN109840841A (en) | A kind of internet finance and money management platform and method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180629 |
|
| RJ01 | Rejection of invention patent application after publication |