CN102866984B - Matrix quantization analytical approach and system thereof in Intelligent Trade - Google Patents

Abstract

The present invention discloses a kind of matrix quantization analytical approach and system of Intelligent Trade, and the method comprises: step 1, and set up the database being used for matrix quantization and analyzing, from database, determine the time cycle, the length in cycle sets by the parameter of matrix; Step 2, for the analysis of historical data, is first add up the matrix with aligned identical order to occur the average probability of ups and downs and average wave amplitude at subsequent time period, then calculates the market risk return rate of different cycles respectively; Step 3, according to the market risk return rate determination trading strategies of matrix current time period; Step 4, according to trading strategies, calls Intelligent Trade program, starts automatically to conclude the business; Step 5, call risk evaluating system to assess the historical transaction record that Intelligent Trade program is done, risk class is provided to the risk of this trading account, and utilizes discrimination model to calculate promise breaking point, then calculated illegal building and the Default Probability of account risk class by KMV model.

Description

Matrix quantization analytical approach and system thereof in Intelligent Trade

Technical field

The present invention relates to the data analysis field of Intelligent Trade, particularly relate to matrix quantization data analysing method and system.

Background technology

Morphological analysis uses many methods in technical Analysis field, is usually divided into two large classes: reversed phase and arrangement form, such as head shoulder type, triple top or the end, dual top or the end, dome and round bottom all belong to reversion phychology, and triangle, rhombus, flag-shaped, rectangle then belongs to arrangement form.But no matter which kind of form, which class form has a problem, is exactly that before not coming out of, who can not conclude and can walk what kind of form out form.So the analysis carried out based on form often needs form by the time could determine trading strategies after coming out of.A kind of way so is had can just to know the probability that he goes up or drops and wave amplitude before form is not come out of? the present invention devises matrix statistical study index just based on this.Artificial intelligence is concluded the business, he mainly studies the mode that the thinking how utilizing computer technology to simulate people solves problem, so, artificial intelligence transaction designs according to the thoughtcast of someone or some people and preset in a program, other people are the thought cannot knowing deviser, also cannot improve this transaction system.The technical matters that this patent will solve is, by matrix statistical study index, artificial intelligence is concluded the business can according to anyone thoughtcast to set transaction system and to improve this transaction system in certain rule.And the key that the parameter designing of rule the invention enables this function to realize just. can say that matrix quantization intelligent trading system is the EA producing EA.(english abbreviation of EA and ExpertAdvisors, Chinese meaning expert advisor, is commonly called as intelligent trading system)

The defect of prior art is:

1, for conventional art analysis, when analyzing historical data, especially concrete data volume is not quantized in morphological analysis, and more or by people experience judges, so accuracy is not high, belongs to fuzzy Judgment, but not quantitative analysis.

2, for conventional art analysis, choose for data the time period scope do not determined, this development for prediction future prospects can not provide clear and definite time range, can only judge, so accuracy is not high according to the experience of people.

3, for the calculating of illegal building, KMV model cannot quantize promise breaking point when applying in financial derivatives transaction.

4、。The data analysis component of each Intelligent Trade is black box, and other people cannot know what the inside is on earth.This is unfavorable for the data analysis component improving black box very much.

Summary of the invention

For solving the problem, the invention provides a kind of matrix quantization data analysing method and system thereof, for making up above technical Analysis defect and realizing real-time management and the prediction of risk, providing effective solution.

The invention discloses matrix quantization analytical approach in a kind of Intelligent Trade, comprising:

Matrix quantization analytical approach in a kind of Intelligent Trade, is characterized in that, comprising:

Step 1, set up the database being used for matrix quantization and analyzing, from database, determine the time cycle, the length of time cycle sets by the parameter of matrix;

Step 2, for the analysis of historical data, first be add up the matrix that previous time period and current time period homography have aligned identical order in the historical data respectively to occur the average probability of ups and downs and average wave amplitude at subsequent time period, then calculate the market risk return rate of previous time period and current time period respectively.User free matrix parameter can decide the time cycle needing statistics, and market risk return rate corresponding to different time cycles can be different;

Step 3, the matrix upper a period of time come out and current time period to aligned identical order provides a kind of algorithm showed with the form of 0,1 combination, so that user calls intelligent automated transaction program according to this combination. verify the validity of this matrix quantization analytic system.

Matrix quantization data analysing method in described Intelligent Trade, described step 1 also comprises:

Step 4, for determining the time cycle, there is the probability that rises or fall and wave amplitude in the next data set of every n data set that upper a period of time and current time period have an aligned identical order;

Step 5, calculate and occur the probability that rises or fall and wave amplitude at the next data set of every n data set with aligned identical order of different time sections, described different time sections is 1 minute, 5 minutes, 15 minutes, the random time section in 30 minutes 1 hour, 4 hours, 1 day, 1 week or January, described n is positive integer.

Matrix quantization data analysing method in described Intelligent Trade, the analysis of described step 2 matrix quantization also comprises:

Step 6, the matrix upper a period of time come out and current time period to aligned identical order provides a kind of algorithm showed with the form of 0,1 combination, so that user calls Intelligent Trade program according to this combination;

If

$\mathrm{Hn}=\frac{1}{n}*\underset{k=0}{\overset{n}{\mathrm{\Σ}}}\left(\mathrm{High}\right[n+2]+\mathrm{High}[n+1]+\mathrm{High}[n\left]\right)/3$

$\mathrm{Ln}=\frac{1}{n}*\underset{k=0}{\overset{n}{\mathrm{\Σ}}}\left(\mathrm{Low}\right[n+2]+\mathrm{Low}[n+1]+\mathrm{Low}[n\left]\right)/3$

$\mathrm{Cn}=\frac{1}{n}*\underset{k=0}{\overset{n}{\mathrm{\Σ}}}\left(\mathrm{Close}\right[n+2]+\mathrm{Close}[n+1]+\mathrm{Close}[n\left]\right)/3$

$\mathrm{Jn}=\frac{1}{n}*\underset{k=0}{\overset{n}{\mathrm{\Σ}}}\mathrm{Hn}*\mathrm{Ln}*\mathrm{Cn}/9$

If Jn+1<Jn, so set zh as 1, if Jn+1>Jn, so set zh as 0;

Hypothesis matrix matrix=n, the number of so zh combination is exactly n, by that analogy; Wherein Hn refers to the flat fare of 3 K line highest prices, Ln refers to the flat fare of 3 K line lowest prices, Cn refers to the flat fare of 3 K line closing prices, Jn refers to 3 K line highest prices, divided by the flat fare of 9 after lowest price is multiplied with closing price, zh refers to that in each combination, each single 0 or 1, High refer to highest price, Low refers to lowest price, and Close refers to closing price;

Step 7, matrix quantization statistical indicator also provides another to generate the algorithm of 0,1 combination: be set to 1 assuming that rise, fall and be set to 0; Hypothesis matrix parameter matrix=n, the number of so zh combination is exactly n, by that analogy; Wherein zh combination represents 0,1 combination that price is formed after matrix conversion, and n is positive integer.

Matrix quantization data analysing method in described Intelligent Trade, described step 2 also comprises:

Step 8, can provide the matrix parameter of any N number of data set in matrix quantization analysis, herein the matrix of an optional 4-9 data set, and described data set is by the highest price of K line, and lowest price, closing price form.

Matrix quantization data analysing method in described Intelligent Trade, described step 3 also comprises:

Step 9, the probability risen occurred according to the next K line of every n the data set with aligned identical order is multiplied by the probability fallen that wave amplitude deducts the next K line appearance of every n the data set with aligned identical order and is multiplied by the risk reward ratio that historical data that wave amplitude obtains runs into this aligned identical data set sequentially, the tendency that the risk reward ratio prediction that user can provide according to market is following, thus determine the strategy of investment;

Step 10, the market risk return rate can calculated according to matrix quantization statistical indicator for the selection of strategy decides, if risk reward ratio is that to do many cost compares little on the occasion of representing market, it is more worthwhile namely to buy in, this value is larger, represents that the adaptive expectations bought in is larger; If risk reward ratio is negative value, representing market, to do empty cost compare little, and it is more worthwhile namely to sell, and negative value is larger, and the adaptive expectations that expression is sold is larger.

Matrix quantization data analysing method in described Intelligent Trade, described step 4 also comprises:

Step 11, discrimination model is nonlinear equation solution, wij=S ^-1(μ _g-μ _b), wherein S ^-1inverse for covariance matrix, μ _gfor target variable is the mean vector of 1; μ _bfor target variable is the mean vector of 0, calculate promise breaking point by utilizing discrimination model;

Step 12, promise breaking point, the net value stability bandwidth of described judged result is accumulated according to user, calculate illegal building DD=(Ea-DP)/Ea* μ a and Default Probability EDF=N (-DD)=1-N (DD), wherein, Ea net value, DP is promise breaking point, μ a is net value stability bandwidth, wherein the calculating of DP promise breaking point make use of discrimination model, and the value of promise breaking point is the scope depending on defective value standard deviation place, and the promise breaking point that different defective value standard deviations is calculated is different.

The present invention also discloses matrix quantization data analysis system in a kind of Intelligent Trade, comprising:

Building database module, sets up the database being used for matrix analysis combination, determines the time cycle from database;

Numerical range module, for for the time cycle, the data assemblies that the matrix quantization that in analysis, a period of time and current time period have aligned identical order is analyzed, user free matrix parameter can decide the data set scope needing statistics;

Statistical module, user can input different values and decide to select the numerical range of which matrix to be used as statistical study, thus gives a forecast to the market risk return rate in future;

The name module of matrix, the matrix providing the upper a period of time of two kinds of algorithms to coming out and current time period to have aligned identical order with the form of 0,1 combination to name this matrix so that user calls according to the title of this combination the validity that Intelligent Trade program carrys out validation matrix quantitative analysis system;

Intelligence automated transaction authentication module, formulates trading strategies according to matrix combination, then calls Intelligent Trade program and start automated transaction, so that evaluating system is according to historical transaction record assessment trading strategies and Strategy Risk.

Discrimination model and KMV model, historical transaction record for doing Intelligent Trade program carries out the calculating of the Default Probability of risk class, first risk class is provided by the risk of risk evaluating system to this trading account, then calculate promise breaking point by discrimination model, then calculate illegal building by KMV model.

The matrix quantization data analysis system of described Intelligent Trade, also comprises:

0,1 combination that the risk reward ratio provided according to matrix quantization data analysis system is corresponding, call Intelligent Trade program Change_mall and can construct out various complex transaction system, each inside Change_mall buys in or sell condition and the condition of closing a position can do "AND", "or" combines, although dealing and the parameter of closing a position often are planted and are only provided 6 parameters, but the investment portfolio binding time cycle can constructed between them but can reach up to ten thousand kinds, wherein Change_mall is the name that Intelligent Trade corresponds to matrix quantization analysis indexes.

The matrix quantization analytic system of described Intelligent Trade, also comprises:

Risk evaluating system, by calling the historical transaction record of intelligent trading system, it (is the patent content (application number is: 201110353391.1)) calling application last year to the assessment of historical transaction record that evaluating system is assessed the transaction record of history automatically herein, but be only limitted to venture worth, the assessment of desired percentage return and risk class.In addition, also supplement the patent of application last year, he calculates illegal building and the Default Probability of account risk class instantly;

The present invention also discloses a kind of Intelligent Trade, and described intelligent trading system is by above-mentioned matrix quantization statistical study index, and intelligent automated transaction program and risk evaluating system form.

Beneficial effect of the present invention is:

1: Intelligent Trade matrix quantization analytic system is a system that can derive numerous Intelligent Trade (EA), can say that he be that to produce below the EA. of EA be my 111100 and 001111 quantification intelligence EA, Fig. 3 done altogether derivative from this system and following table is verify with the discs that pound U.S.A does to report.

As can be seen from report, quantizing the intelligent EA system income of 4 months is 37.55%.And his greateset risk only has 19.8%.Illustrate that this derivative intelligent trading system EA can not only well control risk, and the validity of trading strategies can be verified.

2: the trading strategies after updating finally leaves artificial intelligence transaction (EA) can stablized and make a profit, and the inner parameter of this artificial intelligence transaction (EA) is that others' institute is ignorant.Thus the profit not only protecting investor is secret, also can ensure that intelligent trading system can not excessively use the efficiency causing reducing this system because others usurps, and ensure the space that user can improve at any time simultaneously.

3: quantize transaction for still stranger and advanced domestic most people, Intelligent Trade is due to domestic financial system development imperfection, the people of contact is just less, quantification transaction is the invention enables to become popular, artificial intelligence transaction becomes simplification, and expect that quantizing artificial intelligence transaction becomes following transaction stream and come into vogue, as long as because anyone sets several simple parameter just can realize artificial intelligence quantification transaction.

4: by this system, he achieves the quantification tool of a kind of general production of intelligent transaction.Usually, an Intelligent Trade at least needs two steps, the first, have an idea, the second, this idea of programming realization.And matrix quantization intelligent trading system only needs you to have the first step just can realize Intelligent Trade, in other words this system simplifies the production stage of Intelligent Trade, because you do not need the idea programming to realize you, just can start to carry out Intelligent Trade as long as set several simple parameter.In addition because he also has evaluation system, so he can not only improve the risk identification ability of Intelligent Trade, and enhances the risk control ability of Intelligent Trade.Integrate and see, matrix quantization intelligent trading system achieves the quantification monitoring of dimensions of market fractional analysis from manual transaction to Intelligent Trade and transaction risk.Can say that he is the quantification tool of a kind of general production of intelligent transaction, be the intelligent trading system (EA) of production of intelligent transaction (EA).

Accompanying drawing explanation

Figure 1A-1O is that matrix quantization data analysis index of the present invention introduces block diagram;

Fig. 2 A-2D is that matrix quantization data analysis index of the present invention introduces block diagram;

Fig. 3 is the discs checking report that in beneficial effect of the present invention, pound U.S.A does;

Fig. 4 is the embodiment block diagram that the present invention breaks a contract a little;

Fig. 5 is matrix quantization data analysing method embodiment block diagram of the present invention;

Fig. 6 is matrix quantization data analysing method process flow diagram of the present invention;

Fig. 7 is matrix quantization data analysis system process flow diagram of the present invention.

Embodiment

Provide the specific embodiment of the present invention below, by reference to the accompanying drawings to invention has been detailed description.

Be the introduction of matrix quantization data analysis index as shown in Figure 1A, morphological analysis uses many methods in technical Analysis field, usually two large classes are divided into: reversed phase and arrangement form, such as head shoulder type, triple top or the end, dual top or the end, dome and round bottom all belong to reversion phychology, and triangle, rhombus, flag-shaped, rectangle then belongs to arrangement form.But no matter which kind of form, which class form has a problem, is exactly that before not coming out of, who can not conclude and can walk what kind of form out form.So the transaction carried out based on morphological analysis often needs form by the time could determine trading strategies after coming out of.A kind of way so is had can just to know the probability that he goes up or drops and wave amplitude before form is not come out of? the present invention devises matrix statistical study index just based on this.

For matrix quantization data analysing method, he mainly studies the mode that the thinking how utilizing computer technology to simulate people solves problem, usually, an intelligent trading system designs according to the thoughtcast of someone or some people and preset in a program, other people are the thought cannot knowing deviser, also cannot improve this transaction system.The technical matters that this patent will solve is, by matrix statistical study index, makes matrix quantization data analysing method can according to anyone thoughtcast to set transaction system and to improve this transaction system in certain rule.And the key that the parameter designing of rule the invention enables this function to realize just. can say that matrix quantization intelligent trading system is the EA producing EA.

For the calculating of illegal building, KMV model is to quantize point of breaking a contract when applying in financial derivatives transaction, and the present invention makes the calculating of promise breaking point be achieved by utilizing discrimination model.Thus can accurately calculate at foreign exchange, futures, the illegal building of certain account and Default Probability on other financial derivatives such as stock index.This grading for individual or artificial intelligence transaction system and improve individual or the performance of artificial intelligence transaction system is of great advantage.

Matrix quantization statistical study index is divided into qualitative analysis and quantitative test.

Qualitative analysis:

Cycle, H representative hour wherein, H1 was exactly 1 hour, and M representative minute, M15/30 represents 15 minutes and 30 minutes, and D1, W1 represent sky and week respectively as shown in figure ik.What italicized item represented that MACD calculates falls.What represent that MACD calculates with underscore part rises, and the optimum configurations of MACD is shown in Fig. 1 L.

What bull bear said is take opening price as benchmark, or then calculates the position contrast situation of profit and loss ratio according to 1:3 1:2.

As depicted in figure im, the analysis of matrix position in storehouse has 4 results to position in storehouse, respectively: open a position, adds storehouse, sells shares, close a position.How does that draw these results? see the following form:

Briefly, we deduce according to transaction of taking advantage of a situation, and trend does not become, and yes holds position, trend changes, and is close a position or open a position naturally, if trend does not become, and the direction in trend does not also become, that just should add storehouse, if trend changes, and the direction in trend also changes, and just should sell shares.

How to define trend and direction?

This problem more complicated, first we are defined as megatrend the tendency of sky maps and all figure, within 4 hours, be defined as main trend, the tendency of 1 hour and 4 hours be defined as secondary trend, the tendency of 15 and 30 minutes is defined as direction with the tendency of sky maps.Given by " position in storehouse " four results that everybody sees are the results drawn according to main trend analytical calculation, instead of other trend analysis calculates.So, it is noted that when using if one only see 15 minutes and the short-term of 5 minutes figure objective, so matrix trend studies and judges the position in storehouse analysis that provides to inapplicable you.At this, it is noted that the result that position in storehouse analysis provides is suitable only for do schemes above analysis in 1 hour.So specifically how to define trend and direction? such as: if the average line of 4 hours and 1 hour is all upwards, just think that secondary trend rises, so, during this time 30 minutes and the average line of 15 minutes are just considered to the direction of secondary trend.If the average line of sky maps and 4 hours is all downward, just think that main trend is fallen, so, during this time 4 hours and the average line of 1 hour are just considered to the direction of main trend.Check and regulate and be exactly, if the average line one of week figure and sky maps upwards another one downwards, so we just think that megatrend is checked and regulated, so, the average line of sky maps during this time and 4 hours is just considered to the direction of megatrend.Etc. by that analogy.So, finally judge market be on earth trend or check and regulate time, also introduce forward and reverse concept.

Direct/Reverse is as shown in Fig. 1 N, and forward is exactly: two average lines of two time cycles rise together or fall together.Such as, the average line of 1 hour of secondary trend and 4 hours all gold forks or all extremely pitch, the positive and negative where just display of so secondary trend be forward.Be exactly oppositely: two average lines one of two time cycles are being fallen in the another one that rises.Such as, 4 hours of main trend and the average line of sky maps one dead fork another one gold fork, so positive and negative where just display of main trend is reverse.The value of average line can set inside parameter, is divided into cable release and slow line, and as Fig. 1 L, different average line parameters can cause forward and oppositely occur different results.Specifically depending on situation at that time.

Strategy is as shown in Fig. 1 O, after trend is formed, rising through one section or often will adjust back after drop, power-assisted will be run into during readjustment, be exactly now that we are when opening a position, I was once summarized as four words this situation: trend is hampered and will be hampered existing position by readjustment, and calm need are determined power risk case and shown true qualities.This needs the many observations of user.What touch that top snaps up test is the judgment of deal maker, but the cost of judgment is very little, and the cost of transaction is sometimes high allows you feel surprised.The mistake of transaction is the cost of profit, and misses the cost just judged.So real transaction needs to pay cost.It is exactly the transaction paying cost that chance helps, and is real transaction of taking advantage of a situation.

(be exactly that Double Tops add black clouds caping, Double bottom adds the saturating form of last thorn) that a lot of people of 2B form knows, price will rise and need first fall, and price will be fallen and need first be risen, and sends out first extremely, need to see clearly first chance after so-called.When foreign exchange transaction, do many? do many when rising again up after breaking support by a fall.When do sky, fall again when returning after the power-assisted that rises brokenly and do sky.2B form checks and regulates all the highest trading strategies of success ratio and rate of returns in market.

Quantitative test:

In Figure 1B 0,1 combination be by matrix operation come.The data source of matrix is exactly the opening price of historical quotes data, closing price, highest price and lowest price.Here matrix has only used closing price wherein, highest price and lowest price.Then by certain algorithm, price matrix is changed to the combination only having 0 and 1.Matrix conversion algorithm provides two kinds of algorithms here.

Algorithm 1:

$\mathrm{Hn}=\frac{1}{n}*\underset{k=0}{\overset{n}{\mathrm{\&Sigma;}}}\left(\mathrm{High}\right[n+2]+\mathrm{High}[n+1]+\mathrm{High}[n\left]\right)/3$

$\mathrm{Ln}=\frac{1}{n}*\underset{k=0}{\overset{n}{\mathrm{\&Sigma;}}}\left(\mathrm{Low}\right[n+2]+\mathrm{Low}[n+1]+\mathrm{Low}[n\left]\right)/3$

$\mathrm{Cn}=\frac{1}{n}*\underset{k=0}{\overset{n}{\mathrm{\&Sigma;}}}\left(\mathrm{Close}\right[n+2]+\mathrm{Close}[n+1]+\mathrm{Close}[n\left]\right)/3$

$\mathrm{Jn}=\frac{1}{n}*\underset{k=0}{\overset{n}{\mathrm{\&Sigma;}}}\mathrm{Hn}*\mathrm{Ln}*\mathrm{Cn}/9$

If Jn+1<Jn, so set zh as 1, if Jn+1>Jn, so set zh as 0;

Hypothesis matrix matrix=n, the number of so zh combination is exactly n, by that analogy.

If there is the zh value of continuous five k lines to be, 1001, so this combination is exactly 1001, if there is the zh value of seven k lines to be continuously, 110100, be so combined as: 110100.The rest may be inferred.

Algorithm 2: be set to 1 assuming that rise, falls and is set to 0, if there are continuous five k lines to be respectively, ups and downs are risen, and so this combination is exactly 10111, if there are the ups and downs of six roots of sensation k line to be respectively continuously: fall ups and downs and fall and rise, be so combined as: 010011.The rest may be inferred.Algorithm 2 is special cases of algorithm 1.

Morphological analysis because different forms has different combinations, even if identical form is also mostly different combination, thus cannot according to form statistics instantly or the ups and downs probability in next moment and wave amplitude.Now the matrix by Price Impact is converted to various 0,1 combination by algorithm 1 or algorithm 2.So just can add up have identical 0,1 combination instantly or next time ups and downs probability and wave amplitude.

Go up by average wave amplitude as shown in Figure 1 C, the average wave amplitude of rise of that root K line when the average wave amplitude that goes up refers to rise or the drop in the corresponding at this very moment and now next moment that those those matrixes carved are corresponding.

Rise average probability as shown in figure ip, the rise average probability of that root K line when rise average probability refers to rise or the drop in the corresponding at this very moment and now next moment that those those matrixes carved are corresponding.

Drop average wave amplitude as referring to figure 1e, the average wave amplitude of drop of that root K line when the average wave amplitude that drops refers to rise or the drop in the corresponding at this very moment and now next moment that those those matrixes carved are corresponding.

Drop average probability drops the drop average probability of that root K line of average probability when referring to rise or the drop in the corresponding at this very moment or now next moment that those those matrixes carved are corresponding as shown in fig. 1f.

Risk reward ratio is as shown in Fig. 1 G, 1H, risk reward ratio refers to that corresponding that carve the risk reward ratio of the next root K line of the correspondence of 01 combination that is obtained by matrix algorithms at this very moment or now, and risk reward ratio=rise probability * rise wave amplitude-drop probability * drops wave amplitude.

At this very moment as shown in Figure 1 I, any meaning is represented at this very moment: refer to 01 combination that moment every n root k line is obtained by matrix algorithms at this very moment.

Now that are carved as shown in figure ij, and now what meaning Bi Ke represents: now that carve 01 combination referring to the every n root k of moment instantly line and obtained by matrix algorithms.

The design of the trading strategies of intelligence automated transaction

As Fig. 2 A

1:Buy1 to buy6 refers to the condition bought in.See Fig. 2 D

2:sell1 to sell6 refers to the condition sold.See Fig. 2 D

3:closelong1 to closelong6 refers to flatten and buys in single condition.See Fig. 2 D

4:closeshort1 to closeshort6 refers to flatten and sells single condition.See Fig. 2 D

5:bestlots refers to the hand number of each transaction instantly.See Fig. 2 C

6:max refers to altogether to place an order to be no more than how many hands.See Fig. 2 C

7:matrix be you make the algorithm of matrix of how many k lines, matrix can be arranged from 4 to 9.But through my statistics, find with 6 it is have reasonable effect.If select 9, because the permutation and combination of 9 is really too much, generally our mt4 does not provide enough data to add up, so many times can not calculate out ups and downs probability and wave amplitude.So do not advise setting the combination being matrix with 9.There are enough data words, the matrix combination of more than 10 or 10 can be set.

0,1 combination that the risk reward ratio provided according to matrix quantization analytic system is corresponding, calls matrix intelligent trading system and can construct out various complex transaction system.Each inside Change_mall buys in or sell condition and the condition of closing a position can do "AND", "or" combines.Although dealing and the parameter of closing a position often are planted and only provided 6 parameters, the investment portfolio binding time cycle can constructed between them but can reach up to ten thousand kinds.Wherein Change_mall is the name that Intelligent Trade corresponds to matrix quantization analysis indexes.(see Fig. 2 B)

Optimum configurations rule declaration:

The assignment of 1: from buy1 to buy6, sell1 to sell6, closelong1 to closelong6, closeshort1 to closeshort6 must be all the combination of 0 or 1, such as: 1001,0001111,101010,01101011 etc.The several assignment according to matrix of the combination bit that specifically can set determines.If matrix=4, so the combination of 0,1 more than 4, can not can not be less than 4, by that analogy certainly.Be defaulted as 6.

What 2:buy1 to buy6 set buys in rule, can only set rule of closing a position at closelong1 to closelong6; What sell1 to sell6 set sells rule, can only set rule of closing a position in closeshort1 to closeshort6.Wherein, the rule of buy1 setting does not need certain corresponding closelong1, can set arbitrarily in closelong1 to closelong6, and program can identify corresponding rule of closing a position automatically.

3: each trade variety can set any multi-exchange rule and rule of closing a position, but need at different windowed time, if in the multi-exchange rule of same windowed time, the list of buy1 may have been flattened after meeting the rule of closelong2 because of market, so, do not advise setting trading rules more than more than a kind at same window simultaneously.In addition, the trading rules of buying in, sell or closing a position can set in the different time cycles, EA can identify automatically, there will not be and closed a position after the rule of closing a position of 4 hours period settings occurred in 5 minute cycle, also can not be struck a bargain because the rule set in 1 hours period occurred in 30 minute cycle.

4:bestlots is the hand number at every turn automatically placed an order, from all right between 0.01 to 1000.

5:max refer to allow transaction maximum hand number, when now always conclude the business hand count under the maximum permission reaching setting after singlehanded number, program would not newly increase transaction again, unless there are new single appearance of closing a position.Assignment can between 0.01 to 1000.

6:matirx refers to that you wish to add up with the matrix of much numbers ups and downs probability and the ups and downs wave amplitude of market conditions.If you establish matrix=5, so program opens low receipts to calculate according to the height of every 8 k lines, and statistics has average ups and downs probability corresponding to same matrix and wave amplitude.

Risk evaluating system

After operation after a while, by calling risk evaluating system, analyze the quality of this trading strategies, such as can see the illegal building of trading strategies, Default Probability, risk reward ratio, the size of potential risk, implicit risk number, with this, the indexs such as the ratio of the accounting of irrational transaction and expected yield and venture worth, determine that whether eliminating this trading strategies still improves this trading strategies.

As shown in Figure 4, the most important thing is the parameter of illegal building and Default Probability in intelligent trading system.Illegal building refers to the distance of the next grading of distance, if this value is larger, represent that the possibility of promise breaking is less, Default Probability is also less.Otherwise if this value is less, represent that the possibility of promise breaking is larger, Default Probability is also larger.Such as, if your risk control ability is cited as, illegal building equals 2.1, and so Default Probability is 1.8%, represents that your present wind control ability is very stable, unlikely has the possibility of degradation.Wind control ability is always divided into 5 grades in quantification transaction analysis system, is in extreme danger, dangerous, generally, good, very well.Further, wind control ability might not be decline step by step or rise step by step, and he bypasses the immediate leadership suddenly because of the market of certain secondary burst possibly.But illegal building must be the variation of and then grading and change, and what he pointed to all the time is the illegal building and Default Probability of instantly grading.

First will calculate promise breaking point (illegal building DD=(Ea-DP)/Ea* μ a and Default Probability EDF=N (-DD) when calculating illegal building, wherein, Ea net value, DP is promise breaking point, and μ a is net value stability bandwidth.), but KMV model does not provide the computing formula of promise breaking point when applying in financial derivatives transaction, the present invention utilizes discrimination model solution to obtain here, the value of promise breaking point is the scope depending on defective value standard deviation place, and the promise breaking point that different defective value standard deviations is calculated is different.

Described discrimination model is

The change Sampleweight that sampleweight relates to before point box model substitution solver refers to sample weights, and slover is core algorithm) predictive variable: x, y, z

Predictive variable rough segmentation case: x ₁x ₂x ₃, y ₁y ₂, z ₁z ₂

Calculate wij formula: (wij refers to not sized weight)

wij=S ^-1(μ _G-μ _B)

Wherein S ^-1inverse for covariance matrix, μ _gfor target variable is the mean vector of 1; μ _bfor target variable is the mean vector of 0

Mean value computation formula is as follows:

μ _G=(P _G(x ₁),P _G(x ₂),P _G(x ₃),P _G(y ₁),P _G(y ₂),P _G(z ₁),P _G(z ₂))

μ _B=(P _B(x ₁),P _B(x ₂),P _B(x ₃),P _B(y ₁),P _B(y ₂),P _B(z ₁),P _B(z ₂))

Wherein first rough segmentation case x of first predictive variable x ₁mean value computation formula P _g(x ₁), P _b(x ₁) as follows

Covariance matrix S computing formula is as follows:

The part of different colours calculates the formula difference of covariance, x ₁x ₂x ₃represent three rough segmentation casees of same predictive variable; Y, z represent two other predictive variable (bin is exactly branch mailbox)

Following table light gray part is the variance in a bin

Darker gray part is the covariance in a variable between different bin

Dotted border part is the covariance of bin between different variable

Colourless bold portion is symmetrical with other parts, and namely value is symmetry equivalent

Often kind of different piece covariance formula be all fine or not covariance sum divided by 2 namely

(in formula refer to variance and)

Variance in light gray part covariance matrix S on principal diagonal in bin

$a=\frac{{\mathrm{\&sigma;}}_{G1}^2+{\mathrm{\&sigma;}}_{B1}^2}{2}$

${\mathrm{\&sigma;}}_{G1}^2=P_G\left(x_1\right)-P_G{\left(x_1\right)}^2$

${\mathrm{\&sigma;}}_{B1}^2=P_B\left(x_1\right)-P_B{\left(x_1\right)}^2$

Wherein P _g(x ₁), P _b(x ₁) algorithm the same

Covariance in darker gray part variable between different bin

$b=\frac{{\mathrm{\&sigma;}}_{G1}^2+{\mathrm{\&sigma;}}_{B1}^2}{2}$

${\mathrm{\&sigma;}}_{G1}^2=-P_G\left(x_1\right)\&times;P_G\left(x_2\right)$

${\mathrm{\&sigma;}}_{B1}^2=-P_B\left(x_1\right)\&times;P_B\left(x_2\right)$

The covariance of bin between the different variable of dotted border part

$c=\frac{{\mathrm{\&sigma;}}_{G1}^2+{\mathrm{\&sigma;}}_{B1}^2}{2}$

${\mathrm{\&sigma;}}_{G1}^2=P_G\left(x_1{y}_1\right)-P_G\left(x_1\right)\&times;P_G\left(y_1\right)$

${\mathrm{\&sigma;}}_{B1}^2=P_B\left(x_1{y}_1\right)-P_B\left(x_1\right)\&times;P_B\left(y_1\right)$

Wherein,

As shown in Figure 5, matrix quantization data analysing method embodiment process flow diagram of the present invention, its concrete steps are:

S100, by matrix quantization data analysis index analysis market conditions;

S200, the turnover field rule of design trading strategies and rule of closing a position, select march into the arena rule and the rule that appears on the scene, then perform the combination that S300(returns matrix, both combinations of 0,1);

S300, calls Intelligent Trade program and realizes automated transaction, and allowing computer automatically produce according to the rule of design can for the data analyzed;

S400, by risk evaluating system, the quality of assessment intelligent trading system, performs step S500;

S500, the quality of the rule that analysis matrix quantized data analytic system provides, if also have the space being worth improving, then from the turnover field rule of new design trading strategies and rule of closing a position after improving, and perform S300, no person performs S600;

S600, eliminates the trading strategies that those risks are higher.

That (application number: 201110353391.1), and matrix quantization analytical approach of the present invention, matrix quantization is analyzed intelligent trading system three part and jointly formed by risk evaluation system as Intelligent Trade matrix quantization analytic system.

As shown in Figure 6, the invention discloses matrix quantization analytical approach in a kind of Intelligent Trade, comprising:

Matrix quantization analytical approach in a kind of Intelligent Trade, is characterized in that, comprising:

Step 1, set up the database being used for matrix quantization and analyzing, from database, determine the time cycle, the length of time cycle sets by the parameter of matrix;

Step 2, for the analysis of historical data, first be add up the matrix that previous time period and current time period homography have aligned identical order in the historical data respectively to occur the average probability of ups and downs and average wave amplitude at subsequent time period, then calculate the market risk return rate of previous time period and current time period respectively.User free matrix parameter can decide the time cycle needing statistics, and market risk return rate corresponding to different time cycles can be different;

Step 3, the matrix upper a period of time come out and current time period to aligned identical order provides a kind of algorithm showed with the form of 0,1 combination, so that user calls intelligent automated transaction program according to this combination. verify the validity of this matrix quantization analytic system.

As shown in Figure 7, the present invention also discloses matrix quantization data analysis system in a kind of Intelligent Trade, comprising:

Building database module 10, sets up the database being used for matrix analysis combination, determines the time cycle from database;

Numerical range module 20, for for the time cycle, the data assemblies that the matrix quantization that in analysis, a period of time and current time period have aligned identical order is analyzed, user free matrix parameter can decide the data set scope needing statistics;

Statistical module 30, user can input different values and decide to select the numerical range of which matrix to be used as statistical study, thus gives a forecast to the market risk return rate in future;

The name module 40 of matrix, the matrix providing the upper a period of time of two kinds of algorithms to coming out and current time period to have aligned identical order with the form of 0,1 combination to name this matrix so that user calls Intelligent Trade program according to the title of this combination;

Intelligence automated transaction authentication module 50, formulates trading strategies according to matrix combination, then calls Intelligent Trade program and start automated transaction, so that evaluating system is according to historical transaction record assessment trading strategies and Strategy Risk;

Discrimination model and KMV model 60, historical transaction record for doing Intelligent Trade program carries out the calculating of the Default Probability of risk class, first risk class is provided by the risk of risk evaluating system to this trading account, then calculate promise breaking point by discrimination model, then calculate illegal building by KMV model

The matrix quantization data analysis system of described Intelligent Trade, also comprises:

Computing module, for establishing

$\mathrm{Hn}=\frac{1}{n}*\underset{k=0}{\overset{n}{\mathrm{\&Sigma;}}}\left(\mathrm{High}\right[n+2]+\mathrm{High}[n+1]+\mathrm{High}[n\left]\right)/3$

$\mathrm{Ln}=\frac{1}{n}*\underset{k=0}{\overset{n}{\mathrm{\&Sigma;}}}\left(\mathrm{Low}\right[n+2]+\mathrm{Low}[n+1]+\mathrm{Low}[n\left]\right)/3$

$\mathrm{Cn}=\frac{1}{n}*\underset{k=0}{\overset{n}{\mathrm{\&Sigma;}}}\left(\mathrm{Close}\right[n+2]+\mathrm{Close}[n+1]+\mathrm{Close}[n\left]\right)/3$

$\mathrm{Jn}=\frac{1}{n}*\underset{k=0}{\overset{n}{\mathrm{\&Sigma;}}}\mathrm{Hn}*\mathrm{Ln}*\mathrm{Cn}/9$

If Jn+1<Jn, so set zh as 1, if Jn+1>Jn, so set zh as 0;

Hypothesis matrix matrix=n, the number of so zh combination is exactly n, by that analogy; Wherein Hn refers to the flat fare of 3 K line highest prices, Ln refers to the flat fare of 3 K line lowest prices, Cn refers to the flat fare of 3 K line closing prices, Jn refers to 3 K line highest prices, divided by the flat fare of 9 after lowest price is multiplied with closing price, zh refers to that in each combination, each single 0 or 1, High refer to highest price, Low refers to lowest price, and Close refers to closing price;

Matrix quantization statistical indicator also provides another to generate the algorithm of 0,1 combination: be set to 1 assuming that rise, fall and be set to 0; Hypothesis matrix parameter matrix=n, the number of so zh combination is exactly n, by that analogy; Wherein zh combination represents 0,1 combination that price is formed after matrix conversion, and n is positive integer.

Matrix quantization data analysis system in described Intelligent Trade, described numerical range module also comprises:

Matrix parameter module, for providing the matrix parameter of any N number of data set in matrix quantization analysis, the matrix of an optional 4-9 data set herein, described data set is by the highest price of K line, and lowest price, closing price form.

Matrix quantization data analysis system in described Intelligent Trade, described statistical module also comprises:

Risk-reward module, the probability risen for occurring according to the next K line of every n the data set with aligned identical order is multiplied by the probability fallen that wave amplitude deducts the next K line appearance of every n the data set with aligned identical order and is multiplied by the risk reward ratio that historical data that wave amplitude obtains runs into this aligned identical data set sequentially, the tendency that the risk reward ratio prediction that user can provide according to market is following, thus determine the strategy of investment;

Adaptive expectations module, the matrix quantization statistical indicator of market risk return rate for calculating according to to(for) the selection of strategy decides, if risk reward ratio is that to do many cost compares little on the occasion of representing market, it is more worthwhile namely to buy in, this value is larger, represents that the adaptive expectations bought in is larger; If risk reward ratio is negative value, representing market, to do empty cost compare little, and it is more worthwhile namely to sell, and negative value is larger, and the adaptive expectations that expression is sold is larger.

Matrix quantization data analysis system in described Intelligent Trade, described discrimination model and KMV model also comprise:

Solving promise breaking point, is nonlinear equation solution for discrimination model, wij=S ^-1(μ _g-μ _b), wherein S ^-1inverse for covariance matrix, μ _gfor target variable is the mean vector of 1; μ _bfor target variable is the mean vector of 0, calculate promise breaking point by utilizing discrimination model;

Calculate Default Probability module, for accumulating promise breaking point, the net value stability bandwidth of described judged result according to user, calculate illegal building DD=(Ea-DP)/Ea* μ a and Default Probability EDF=N (-DD)=1-N (DD), wherein, Ea net value, DP is promise breaking point, μ a is net value stability bandwidth, wherein the calculating of DP promise breaking point make use of discrimination model, the value of promise breaking point is the scope depending on defective value standard deviation place, and the promise breaking point that different defective value standard deviations is calculated is different.

The matrix quantization data analysis system of described Intelligent Trade, also comprises:

0,1 combination that the risk reward ratio provided according to matrix quantization data analysis system is corresponding, call Intelligent Trade program Change_mall and can construct out various complex transaction system, each inside Change_mall buys in or sell condition and the condition of closing a position can do "AND", "or" combines, although dealing and the parameter of closing a position often are planted and are only provided 6 parameters, but the investment portfolio binding time cycle can constructed between them but can reach up to ten thousand kinds, wherein Change_mall is the name that Intelligent Trade corresponds to matrix quantization analysis indexes.

The matrix quantization analytic system of described Intelligent Trade, also comprises:

Risk evaluating system, by calling the historical transaction record of intelligent trading system, it (is the patent content (application number is: 201110353391.1)) calling application last year to the assessment of historical transaction record that evaluating system is assessed the transaction record of history automatically herein, but be only limitted to venture worth, the assessment of desired percentage return and risk class.Supplement the patent of application last year, he calculates illegal building and the Default Probability of account risk class instantly herein;

The present invention also discloses a kind of Intelligent Trade, and described intelligent trading system is by above-mentioned matrix quantization statistical study index, and Intelligent Trade program and risk evaluating system form.

Those skilled in the art, under the condition not departing from the spirit and scope of the present invention that claims are determined, can also carry out various amendment to above content.Therefore scope of the present invention is not limited in above explanation, but determined by the scope of claims.

Claims (9)

1. a matrix quantization analytical approach in Intelligent Trade, is characterized in that, comprising:

Step 1, set up the database being used for matrix quantization and analyzing, from database, determine the time cycle, the length of time cycle is set by the parameter of matrix;

Step 2, for the analysis of historical data, first be add up the matrix that previous time period and current time period homography have aligned identical order in the historical data respectively to occur the average probability of ups and downs and average wave amplitude at subsequent time period, then the market risk return rate of previous time period and current time period is calculated respectively, the free matrix parameter of user decides the time cycle needing statistics, and market risk return rate corresponding to different time cycles can be different;

Step 3, the matrix upper a period of time come out and current time period to aligned identical order provides a kind of algorithm showed with the form of 0,1 combination, so that user calls intelligent automated transaction program according to this combination, verify the validity of this matrix quantization analytic system.

2. matrix quantization analytical approach in Intelligent Trade as claimed in claim 1, is characterized in that, also comprise:

Step 4, for determining the time cycle, there is the probability that rises or fall and wave amplitude in the next data set of every n data set that statistically a period of time and current time period have an aligned identical order;

Step 5, and add up and occur the probability that rises or fall and wave amplitude at the next data set of every n data set with aligned identical order of different time sections, described different time sections is 1 minute, 5 minutes, 15 minutes, 30 minutes, the random time section in 1 hour, 4 hours, 1 day, 1 week or January, described n was positive integer.

3. matrix quantization analytical approach in Intelligent Trade as claimed in claim 1, is characterized in that, also comprise:

Step 6, the matrix upper a period of time come out and current time period to aligned identical order provides a kind of algorithm showed with the form of 0,1 combination, so that user calls Intelligent Trade program according to this combination;

If

$H_n=\frac{1}{n}*\underset{k=0}{\overset{n}{\&Sigma;}}(High\&lsqb;k+2\&rsqb;+High\&lsqb;k+1\&rsqb;+High\&lsqb;k\&rsqb;)/3$

$L_n=\frac{1}{n}*\underset{k=0}{\overset{n}{\&Sigma;}}(Low\&lsqb;k+2\&rsqb;+Low\&lsqb;k+1\&rsqb;+Low\&lsqb;k\&rsqb;)/3$

$C_n=\frac{1}{n}*\underset{k=0}{\overset{n}{\&Sigma;}}(Close\&lsqb;k+2\&rsqb;+Close\&lsqb;k+1\&rsqb;+Close\&lsqb;k\&rsqb;)/3$

$Jn=\frac{1}{n}*\underset{k=0}{\overset{n}{\&Sigma;}}Hn*Ln*Cn/9$

If Jn+1<Jn, so set zh as 1, if Jn+1>Jn, so set zh as 0;

Hypothesis matrix matrix=n, the number of so zh combination is exactly n, by that analogy; Wherein Hn refers to the flat fare of 3 K line highest prices, Ln refers to the flat fare of 3 K line lowest prices, Cn refers to the flat fare of 3 K line closing prices, Jn refers to 3 K line highest prices, and divided by the flat fare of 9 after lowest price is multiplied with closing price, zh to refer in each combination each single 0 or 1, High refers to highest price, Low refers to lowest price, and Close refers to closing price, and k refers to natural number;

Step 7, matrix quantization statistical indicator also provides another to generate the algorithm of 0,1 combination: be set to 1 assuming that rise, fall and be set to 0; Hypothesis matrix parameter matrix=n, the number of so zh combination is exactly n, by that analogy; Wherein zh combination represents 0,1 combination that price is formed after matrix conversion, and n is positive integer.

4. matrix quantization analytical approach in Intelligent Trade as claimed in claim 3, it is characterized in that, described step 2 also comprises:

Step 8, there is provided the matrix parameter of any N number of data set in matrix quantization analysis, select the matrix of 4-9 data set herein, N is from 1 to just infinite any positive integer, described data set is by the highest price of K line, and what lowest price, closing price were calculated by step 70,1 combines.

5. matrix quantization analytical approach in Intelligent Trade as claimed in claim 1, is characterized in that, also comprise:

Step 9, the probability risen occurred according to the next K line of every n the data set with aligned identical order is multiplied by the probability fallen that wave amplitude deducts the next K line appearance of every n the data set with aligned identical order and is multiplied by the risk reward ratio that historical data that wave amplitude obtains runs into this aligned identical data set sequentially, the tendency that the risk reward ratio prediction that user provides according to market is following, thus determine the strategy of investment;

Step 10, the market risk return rate calculated according to matrix quantization statistical indicator for the selection of strategy decides, if risk reward ratio is that to do many cost compares little on the occasion of representing market, it is more worthwhile namely to buy in, this value is larger, represents that the adaptive expectations bought in is larger; If risk reward ratio is negative value, representing market, to do empty cost compare little, and it is more worthwhile namely to sell, and negative value is larger, and the adaptive expectations that expression is sold is larger.

6. matrix quantization analytical approach in Intelligent Trade as claimed in claim 1, it is characterized in that, described step 3 also comprises:

Step 11, discrimination model is nonlinear equation solution, wij=S ^-1(μ _g-μ _b), wherein S ^-1inverse for covariance matrix, μ _gfor target variable is the mean vector of 1; μ _bfor target variable is the mean vector of 0, calculate promise breaking point by utilizing discrimination model;

Step 12, according to promise breaking point, the net value stability bandwidth of the judged result of user's accumulation, calculate illegal building DD=(Ea-DP)/Ea* μ a and Default Probability EDF=N (-DD)=1-N (DD), wherein, Ea net value, DP is promise breaking point, μ a is net value stability bandwidth, wherein the calculating of DP promise breaking point make use of discrimination model, and the value of promise breaking point is the scope depending on defective value standard deviation place, and the promise breaking point that different defective value standard deviations is calculated is different.

7. a matrix quantization data analysis system in Intelligent Trade, is characterized in that, comprising:

Building database module, sets up the database being used for matrix analysis combination, determines the time cycle from database;

Numerical range module, for for the time cycle, the data assemblies that the matrix quantization that in analysis, a period of time and current time period have aligned identical order is analyzed, the free matrix parameter of user decides the data set scope needing statistics;

Statistical module, user inputs different values and decides to select the numerical range of which matrix to be used as statistical study, thus gives a forecast to the market risk return rate in future;

Matrix combination name module, the matrix providing the upper a period of time of two kinds of algorithms to coming out and current time period to have aligned identical order with the form of 0,1 combination to name this matrix so that user calls according to the title of this combination the validity that Intelligent Trade program carrys out validation matrix quantitative analysis system;

Intelligence automated transaction authentication module, formulates trading strategies according to matrix combination, then calls Intelligent Trade program and start automated transaction, so that evaluating system is according to historical transaction record assessment trading strategies and Strategy Risk;

Discrimination model and KMV model, historical transaction record for doing Intelligent Trade program carries out the calculating of the Default Probability of risk class, first risk class is provided by the risk of risk evaluating system to this trading account, then calculate promise breaking point by discrimination model, then calculate illegal building by KMV model.

8. matrix quantization data analysis system in Intelligent Trade as claimed in claim 7, is characterized in that, also comprise:

0,1 combination that the risk reward ratio provided according to matrix quantization data analysis system is corresponding, call Intelligent Trade program Change_mall and construct out various complex transaction system, each inside Change_mall buys in or sell condition and the condition of closing a position can do "AND", "or" combines, although dealing and the parameter of closing a position often are planted and are only provided 6 parameters, but the investment portfolio binding time cycle can constructed between them but can reach up to ten thousand kinds, wherein Change_mall is the name that Intelligent Trade corresponds to matrix quantization analysis indexes.

9. matrix quantization data analysis system in Intelligent Trade as claimed in claim 7, is characterized in that, also comprise:

Risk evaluating system, by calling the historical transaction record of intelligent trading system, evaluating system is assessed the transaction record of history automatically, and assesses the risk class of account instantly, the illegal building of this risk class for confirmation and Default Probability.