CN108921688A - Construct the method and device of prediction model - Google Patents
Construct the method and device of prediction model Download PDFInfo
- Publication number
- CN108921688A CN108921688A CN201810707589.7A CN201810707589A CN108921688A CN 108921688 A CN108921688 A CN 108921688A CN 201810707589 A CN201810707589 A CN 201810707589A CN 108921688 A CN108921688 A CN 108921688A
- Authority
- CN
- China
- Prior art keywords
- variable
- sequence
- variables
- variables sequence
- model
- 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
- 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
- G06Q40/02—Banking, e.g. interest calculation or account maintenance
-
- 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
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Engineering & Computer Science (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
This specification embodiment provides a kind of method and apparatus that building prediction model is used to predict the liquidity scale, according to this method, it obtains first with the liquidity scale to be predicted there are the Variables Sequence that the variable of certain incidence relation is constituted, such Variables Sequence is to be sampled and obtained to the variable of multiple historical periods.Then ensure that each Variables Sequence is stationary sequence by stationarity detection, and there are causalnexus with the liquidity scale for the variable for ensuring to be related to by Granger CaFpngerusality test.On this basis, the variable of same historical period and the liquidity scale are arranged into the column vector for the period, the column vector of each historical period is input in vector auto regression VAR model.By being trained with the column vector of multiple historical periods, the parameter matrix and disturbance term in vector auto regression VAR model are determined, and the VAR model of parameter will be determined as prediction model.
Description
Technical field
This specification one or more embodiment is related to field of computer technology, more particularly to the side of building prediction model
Method and device.
Background technique
The appearance of internet finance greatly have stimulated the business development of the traditional mechanisms such as bank, various innovative finance
Business continues to bring out, and bigger challenge is brought to management of liquidity.In management of liquidity, the management of liquidity risk is extremely
It closes important.For financial institution, itself illiquidity can cause liquidity risk even crisis, and excessive flowing
Property then means that efficiency in the use of funds lowly leads to lost revenue.For management organization, such as Central Bank, one of management objectives
I.e. regulating the market is expected, and the fluctuation of level out market mobility avoids the occurrence of systemic liquidity risk.
In some schemes, certain methods are proposed to calculate the liquidity scale, market fund is reflected with the liquidity scale
Mobility.For example, Bank of Communications's financial research center proposes Bank of Communications Inter-Bank Market slamp value IBLI, again
Referred to as hand over Yin Aibuli index.The various transactional operations that the calculation method of these slamp values is all based on past period become
Amount cannot but predict following mobility to portray the financial liquidity of that period.In general, the liquidity scale
Calculating often relates to many variables, and usually interacts between these variables and slamp value.Such as capital interest rate
It will affect market financial liquidity, and market financial liquidity can also influence capital interest rate in turn.Such relationship to flow
The prediction more difficult of dynamic property index.
Accordingly, it would be desirable to there is improved plan, the liquidity scale is more efficiently predicted, thus preferably management flowing
Property risk.
Summary of the invention
This specification one or more embodiment describes a kind of method for constructing prediction model, and this method is based on vector
Autoregression model VAR constructs the prediction model for predicting the liquidity scale, thus effectively to following the liquidity scale into
Row Predict and analysis.
According in a first aspect, providing a kind of method for constructing prediction model, the prediction model is commutative for predicting
The liquidity scale of resource, the method includes：
Obtain the first Variables Sequence, the second Variables Sequence and index series, first Variables Sequence and the second variable
Sequence is respectively to carry out sampling shape to the first variable and the second variable respectively in the multiple historical periods being sequentially arranged
At Variables Sequence, wherein first variable is associated with the scale of operation of the commutative resource, second variable with
The return rate of the commutative resource is associated；The index series includes, the corresponding flowing of the multiple historical period
Property index；
The stationarity of the first Variables Sequence, the second Variables Sequence and index series is detected, is become to obtain stable first
Measure sequence, the second Variables Sequence and index series；
Based on stable first Variables Sequence, the second Variables Sequence and index series, determine that each historical period is corresponding
Column vector, wherein the corresponding column vector Yi of the i-th historical period includes, the first variable, the second variable of i-th historical period, with
And the liquidity scale of i-th historical period；
By the corresponding column vector input vector autoregression VAR model of each historical period, the mould of VAR model is determined
Shape parameter, thus using the VAR model comprising above-mentioned model parameter as the prediction model.
In one embodiment, the first variable includes one or more of following variable：Reverse back purchase, medium term lending is just
Sharp MLF, ultra-short term reverse back purchase SLO, and the convenient SLF of standing debt-credit mortgages supplementary loan PSL；
Second variable includes one or more groups of in following variable：Repo rate and trading volume；Inter-bank rate and friendship
Yi Liang；Interest rate swap interest rate and trading volume；Yield of public debt and trading volume.
According to a kind of possible design, using unit root test, come detect the first Variables Sequence, the second Variables Sequence and
The sequence of unit root will be present as unstable sequence in the stationarity of index series.
In a kind of possible embodiment, in detecting the first Variables Sequence, the second Variables Sequence and index series
In the jiggly situation of any sequence, calculus of differences is carried out to jiggly sequence, and test again, until obtaining
To stable first Variables Sequence, the second Variables Sequence and index series.
In one embodiment, method further includes：To the first Variables Sequence/second Variables Sequence progress and index series
Granger causality test, to examine the causal influence relationship of the first variable/between the second variable and the liquidity scale.
Specifically, in one embodiment, by Granger causality test, it is ensured that the first variable and the second variable
In at least one variable and the liquidity scale have Granger two-way causal relationship.
In a kind of possible design, Granger causality test is carried out in the following manner：Determine lag order m；
Based on lag order, the first recurrence processing is carried out to the lag item of the liquidity scale, returns processing result to first based on first
The bivariate lag item of variable/the carries out the second recurrence processing, obtains statistic；It is examined based on the statistic using F to examine
Test the first variable/second variable whether with the liquidity scale there are Granger causalities.
According to a kind of embodiment, VAR model is used for, for lag order m, the column based on continuous m historical period to
Amount, corresponding m parameter matrix and disturbance term determine next historical period after the continuous m historical period
Corresponding column vector；Correspondingly, the model parameter for determining VAR model includes：Determine lag order m；And based on described each
The corresponding column vector of historical period determines the m parameter matrix and the disturbance term.
According to a kind of possible design, lag order m is determined in the following manner：Examine determination stagnant using likelihood ratio LR
Order m afterwards；Or according to akaike information criterion AIC, determine lag order m；Or according to Schwarz information criterion SC, really
Determine lag order m.
In one embodiment, method further includes carrying out pulse response analysis for the VAR model, determining the first change
The/the second variable is measured to the pulse effects of the liquidity scale.
According to second aspect, a kind of device for constructing prediction model is provided, the prediction model is for predicting commutative money
The liquidity scale in source, described device include：
Retrieval unit, is configured to obtain the first Variables Sequence, the second Variables Sequence and index series, and described the
One Variables Sequence and the second Variables Sequence are respectively, in the multiple historical periods being sequentially arranged respectively to the first variable
The Variables Sequence that sampling formation is carried out with the second variable, wherein the scale of operation of first variable and the commutative resource
Associated, second variable is associated with the return rate of the commutative resource；The index series includes, the multiple
The corresponding the liquidity scale of historical period；
Stationarity detection unit is configured to the steady of the first Variables Sequence of detection, the second Variables Sequence and index series
Property, to obtain stable first Variables Sequence, the second Variables Sequence and index series；
Vector determination unit is configured to stable first Variables Sequence, the second Variables Sequence and index series, really
Determine the corresponding column vector of each historical period, wherein the corresponding column vector Yi of the i-th historical period includes, i-th historical period
The liquidity scale of first variable, the second variable and i-th historical period；
Parameter determination unit is configured to the corresponding column vector input vector autoregression VAR mould of each historical period
Type determines the model parameter of VAR model, thus using the VAR model comprising above-mentioned model parameter as the prediction model.
According to the third aspect, a kind of computer readable storage medium is provided, computer program is stored thereon with, works as institute
When stating computer program and executing in a computer, the method that enables computer execute first aspect.
According to fourth aspect, a kind of calculating equipment, including memory and processor are provided, which is characterized in that described to deposit
It is stored with executable code in reservoir, when the processor executes the executable code, the method for realizing first aspect.
The method and apparatus provided by this specification embodiment, the vector auto regression of selection more closing to reality application
VAR model constructs prediction model, to realize the Predict to the liquidity scale of commutative resource.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, making required in being described below to embodiment
Attached drawing is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, right
For those of ordinary skill in the art, without creative efforts, it can also be obtained according to these attached drawings
Its attached drawing.
Fig. 1 shows the implement scene schematic diagram of one embodiment of this specification disclosure；
Fig. 2 shows the method flow diagrams for constructing prediction model according to one embodiment；
Fig. 3 shows the schematic block diagram of the model construction device according to one embodiment.
Specific embodiment
With reference to the accompanying drawing, the scheme provided this specification is described.
Fig. 1 is the implement scene schematic diagram of one embodiment that this specification discloses.In Fig. 1, computing platform be can be
It is any have calculate, the equipment of processing capacity or device clusters, for construct predict commutative resource the liquidity scale it is pre-
Survey model.For example, computing platform can be the server of financial institution, or provide the server of financial product, such as " ant
Ant wealth " server.The server can be used for constructing the prediction model of prediction Inter-Bank Market the liquidity scale.
In order to construct prediction model, computing platform is obtained to exist with the liquidity scale to be predicted first and is centainly associated with
The Variables Sequence that the variable of system is constituted, such Variables Sequence is to be sampled and obtained to the variable of multiple historical periods.One
As, the Variables Sequence to be obtained may include the Variables Sequence of a class variable associated with resource operation scale (in Fig. 1
It is shown as scale of operation class variable sequence), and the Variables Sequence of two class variables associated with resource return rate (shows in Fig. 2
It is out interest rate class variable sequence).Then ensure that each Variables Sequence is stationary sequence, and the variable being related to refers to mobility
There are causalnexus for mark.On this basis, a class variable of same historical period, two class variables and the liquidity scale are arranged
For the column vector of the period, the column vector of each historical period is input in vector auto regression VAR model.By with multiple
The column vector of historical period is trained, and determines parameter matrix and disturbance term in vector auto regression VAR model, and will determine
The VAR model of parameter is as prediction model out.Further, impulse response point can also be carried out to the VAR model determined
Analysis, to determine the influence of a class variable and/or two class variables to the liquidity scale.The VAR model so determined can be used for
The liquidity scale of future time period is effectively predicted.The specific implementation process of the above design is described below.
Fig. 2 shows the method flow diagrams for constructing prediction model according to one embodiment；This method can be had by any
Calculating, the device of processing capacity, equipment, platform, device clusters execute, such as computing platform shown in Fig. 1, and the calculating is flat
Platform specifically can be the server of such as ant wealth.As shown in Fig. 2, this method at least includes the following steps：Step 21, it obtains
Take the first Variables Sequence, the second Variables Sequence and index series, first Variables Sequence and the second Variables Sequence difference
To carry out the variable of sampling formation to the first variable and the second variable respectively in the multiple historical periods being sequentially arranged
Sequence, the index series include, the corresponding the liquidity scale of the multiple historical period；Step 23, first is detected
The stationarity of Variables Sequence, the second Variables Sequence and index series, to obtain stable first Variables Sequence, the second variable
Sequence and index series；Step 25, the corresponding column vector of each historical period is determined, wherein the column vector Yi of the i-th historical period
The liquidity scale of the first variable, the second variable and i-th historical period including, i-th historical period；Step 27,
By the column vector input vector autoregression VAR model of each historical period, the model parameter of VAR model is determined, thus will
VAR model comprising above-mentioned model parameter is as the prediction model.The specific of each step in the above method is described below
Executive mode.
It is appreciated that the prediction model to be constructed is used to predict the liquidity scale of commutative resource.Commutative resource can
Be it is various can be most typical in market with the resource of certain price free exchange, commutative resource is currency.However, with
The release of various financial products, commutative resource can also include that various ideal moneys are derived based on real money
Quasi-money spin-off, etc..It is described by taking currency as an example below, but embodiment can be generalized to other applicable possibility
Commutative resource.
In order to construct prediction model, first in step 21, the first Variables Sequence, the second Variables Sequence and index are obtained
Sequence.
First variable is that there are the exogenous variables of incidence relation with the liquidity scale to be predicted.In one embodiment,
First variable is variable associated with the scale of operation of commutative resource, such as associated with Central Bank's currency operations scale
Variable.More specifically, in one embodiment, the first variable includes one or more of following variable：Reverse back purchase, mid-term
Convenient MLF is borrowed or lent money, ultra-short term reverse back purchases SLO, standing debt-credit convenient SLF, mortgage supplementary loan PSL, etc..Correspondingly, first
Variables Sequence is to carry out the variable sequence of sampling formation to the first variable respectively in the multiple historical periods being sequentially arranged
Column.Above-mentioned historical period can be hour, half a day, one day, one week or even one month etc. any scheduled period.Than
More typical, most bank's relationship tradings are all settled accounts as unit of day, therefore, in one embodiment, when above-mentioned history
Duan Weitian.In following example, it was illustrated and described with one day for a historical period.However, it should be understood that according to can
The billing cycle of exchange resource and the difference of variable, historical period are possible to take different duration units.It was gone through at one day for one
In the example of history period, the first Variables Sequence is, interim in phase of history, samples according to the date to the first variable,
And the Variables Sequence formed according to date collected sequence.For example, the first Variables Sequence can be expressed as, A=(at,at+1,
at+2..., at+n), wherein atIndicate the value that t some day samples variable a in the past, at+1Expression t+1 days to variable a sampling
Value, at+2Indicate the t+2 days values, etc. to variable a sampling.It is appreciated that the first variable may include multiple variables, such as before
The reverse back purchase stated, the convenient MLF of medium term lending, ultra-short term reverse back purchase SLO etc..At this point, the first Variables Sequence may include to each
A first variable carries out the subsequence of sequential sampling formation.
On the other hand, the second variable is also that there are the variables of incidence relation with the liquidity scale to be predicted.In a reality
Apply in example, the second variable is variable associated with the return rate of commutative resource, such as with each time limit fund of Inter-Bank Market
The associated variable of interest rate index.More specifically, in one embodiment, second variable includes one in following variable
Group or multiple groups：Repo rate and trading volume；Inter-bank rate and trading volume；Interest rate swap interest rate and trading volume；National debt income
Rate and trading volume.Correspondingly, the second Variables Sequence is to become respectively to second in the multiple historical periods being sequentially arranged
Amount carries out the Variables Sequence of sampling formation.As previously mentioned, historical period can be any scheduled period, but needs and adopt
The historical period for collecting the first Variables Sequence is consistent.Than more typical, above-mentioned historical period is day.Then, the second Variables Sequence
It can be expressed as, B=(bt,bt+1,bt+2..., bt+n), wherein btIndicate the value that t some day samples variable b in the past, bt+1Table
Show the t+1 days values to variable b sampling, bt+2Indicate the t+2 days values, etc. to variable b sampling.More specifically, the second variable can
Comprising being grouped variable, such as one group of variable of repo rate and corresponding trading volume composition.At this point, the b in the second Variables Sequencet+i
It may include bt+i=(brt+i, bvt+i), wherein brt+iFor t+i days repurchase rates, bvt+iFor corresponding to t+i days repurchase rates
Trading volume.It is appreciated that the second variable may include multiple groups variable, such as repo rate above-mentioned and trading volume；Inter-bank lending and borrowing
Interest rate and trading volume etc..At this point, the second Variables Sequence may include the son for carrying out sequential sampling formation to the second variable of each group
Sequence.
In addition, also obtaining index series, that is, the mobility of the corresponding commutative resource of multiple historical periods refers to
Mark the sequence constituted.The liquidity scale can be known for measuring the data target of the mobility of commutative resource.Example
Such as, in order to reflect the mobility of market fund, Bank of Communications's financial research center proposes the flowing of Bank of Communications's Inter-Bank Market
Sex index IBLI, also known as friendship Yin Aibuli index.IBLI index can be used as the liquidity scale here.Below just with
It is described for IBLI index.It is to be understood that for other commutative resources, it can be using other mobility
Index is measured；Or for market fund, it is also possible to be referred to using the index that other mathematical methods determine as mobility
Mark, is not limited thereto.
For above-mentioned the liquidity scale, need to obtain identical as the historical period that the first variable and the second variable acquire
The corresponding the liquidity scale of historical period, thus composing indexes sequence.
On the basis of obtaining the first Variables Sequence, the second Variables Sequence and index series respectively, in step 23, detection
The stationarity of first Variables Sequence, the second Variables Sequence and index series, to obtain stable first Variables Sequence, second
Variables Sequence and index series.
Data sequence is steadily subsequent builds and the basis using vector auto regression VAR model, therefore, in the step
In, stationary test is carried out to each sequence.If detecting that some Variables Sequence is unstable, stationarity processing can be carried out,
Such as difference, it then tests again, until obtaining stable Variables Sequence.
It can be using a variety of methods come the stationarity of inspection data sequence.
In one embodiment, using unit root test, to detect the stationarity of each sequence.If certain Variables Sequence
In there are unit roots, then it is assumed that the sequence is jiggly time series.
The general thought of unit root test is as follows.Define Random time sequence { Xt, t=1,2 ..., wherein Xt=ρ Xt-1+
εt, definition is rewritten as following form：(1- ρ L)=ε, wherein L is lag operator, and 1- ρ L is lag operator multinomial,
Characteristic equation is 1- ρ z=0, there is root z=1/p.Work as ρ<When 1, { XtIt is stationary sequence.As ρ=1, there are one for time series
A unit root, at this time { XtIt is a unit root process, it is also known as singly had suffered journey, is unstable sequence, but passes through first difference
/ after can become stationary sequence.And as ρ > 1, { XtIt is a kind of non-stationary process with so-called explosion root, pass through
Non-stationary process is remained as after first difference.
Unit root test can be carried out based on a variety of statistics, such as based on DF (Dickey Fuller) statistic.?
In the case where carrying out unit root test based on DF statistic, for Variables Sequence Xt=ρ Xt-1+ε；
Its null hypothesis and alternative hypothesis are respectively：
H0：ρ=1, XtNon-stationary；
H1：ρ<1, XtSteadily.
Under null hypothesis establishment condition, DF statistic is calculated:
If DF is greater than critical value, receive H0, it is believed that XtNon-stationary；If DF is less than or equal to critical value, refuses H0, recognize
For XtSteadily.
In addition to above be based on DF statistic carry out unit root test, be also based on such as ADF (new-added item DF) statistic or
Other statistics carry out unit root test.
In other embodiments, can also test variable sequence by other means stationarity, such as test variable sequence
Mean value, variance and covariance of column etc..
It determines that some Variables Sequence is unstable if examined, tranquilization processing can be carried out to it.Implement at one
In example, difference processing operation is carried out to jiggly Variables Sequence.In other embodiments, some other calculations can also be used
Method carries out tranquilization processing, such as region average processing, the processing of section median, accumulative summation process, section smoothing processing
Deng.By tranquilization processing and then secondary carry out stationary test, until obtaining stable Variables Sequence.
In this way, obtaining stable Variables Sequence by step 23.
The corresponding column vector of each historical period is determined in step 25 based on stable Variables Sequence, wherein the i-th history
The column vector Yi of period includes the flowing of the first variable, the second variable and i-th historical period of i-th historical period
Property index.It is appreciated that the first Variables Sequence, the second Variables Sequence and index series be all arrange sequentially in time when
Between sequence.Time it will be considered as transverse dimensions, then each Variables Sequence can correspond to the row vector of a transverse direction.By
After one Variables Sequence, the second Variables Sequence and index series time unifying, different variables under available same historical period
The column vector of composition.
For example, if each Variables Sequence uses following presentation：
First Variables Sequence A=(at,at+1,at+2..., at+n)；
Second Variables Sequence B=(bt,bt+1,bt+2..., bt+n)；
Index series C=(ct,ct+1,ct+2..., ct+n),
So, each variable under same historical period may be constructed the column vector of the historical period, for example, Yt=(at,
bt,ct),Yt+1=(at+1,bt+1,ct+1) ..., Yt+i=(at+i,bt+i,ct+i)。
As previously mentioned, the first variable and the second variable can include multiple variables, correspondingly, the first Variables Sequence and
Second Variables Sequence respectively contains multiple subsequences.For example, the first variable a includes a1, a2 and a3, the second variable include b1 and
B2, correspondingly, the corresponding column vector of same historical period can be expressed as： Yt+i=(a1t+i,a2t+i, a3t+i, b1t+i,b2t+i,
ct+i).In this way, the column vector of available each historical period.
Based on the corresponding column vector of each historical period so got, the building of prediction model can be carried out, especially
It is the building of vector auto regression VAR model.
In one embodiment, before constructing VAR model, also to the first variable and the second variable carry out Granger because
Fruit relational checking, to examine the causal influence relationship between the first variable, the second variable and the liquidity scale.
Granger causality test was by Nobel prize in economics winner Clive Granger (Clive in 2003
W.J.Granger it) proposes first, for the causality between analysing economic variables.Under time series situation, two warps
Granger causality between Ji variable X, Y is defined as：If under conditions of containing the past information of variable X, Y, to change
The prediction effect of amount Y is better than the prediction effect only individually carried out by the past information of Y to Y, i.e. variable X helps to explain and become
Measure the variation in future of Y, then it is assumed that variable X is the granger cause of induced variable Y.
Mathematically, it is assumed that variable X, the relationship between Y pass through following formula and express：
If α is integrally not zero, and λ generally zero, it is considered that X has unidirectional influence to Y；
If λ is integrally not zero, and generally zero, it is considered that Y has unidirectional influence to X；
If α and λ are integrally not zero, it is considered that Y is with X, there are two-way influences；
If α and λ generally zero, it is considered that there is no influence by Y and X.
It in one embodiment, can be first by current Y to all in order to carry out Granger causality test
Lag item Y returns and excludes the lag item of X, i.e. lag item Y of the Y to Yt-1, Yt-2..., Yt-mAnd its dependent variable returns, and obtains
To controlled residual sum of squares (RSS) RSSR；Then the recurrence containing lag item X is done, i.e., is added in regression equation in front
Item X is lagged, unconfined residual sum of squares (RSS) RSSUR is obtained.At null hypothesis H0：α generally zero, α 1=α 2=...=α m=
0, i.e. lag item X is not belonging to this recurrence.Then it is checked whether to receive the null hypothesis with F inspection.If the F value calculated is more than
Critical value F α, then refuse null hypothesis, then lag item X belongs to this recurrence, shows that X is the granger cause of Y.Exchange X, Y
Set, can examine Y whether the granger cause for being X.
In the above manner, the first variable a and the liquidity scale c can be examined to whether there is causal influence relationship respectively,
And second variable c and the liquidity scale c whether there is causal influence relationship.It is similar with aforementioned process, target is become first
Amount is that the lag item of the liquidity scale carries out the first recurrence processing, obtains controlled residual sum of squares (RSS) RSSR；It is then based on
One, which returns processing result, carries out the second recurrence processing to the lag item of first the second variable of variable a/ b, obtains statistic.Then,
Based on the statistic examined using F inspection first the second variable of variable a/ b whether with the liquidity scale there are Granger because
Fruit relationship.
So, it is ensured that the first variable and the second variable being introduced into subsequent vector auto regression VAR model, be all with
The liquidity scale has the variable of causal influence relationship.In one embodiment, by Granger causality test, it is ensured that
At least one variable and the liquidity scale in first variable and the second variable have two-way causal influence relationship.
In the above checkout procedure, lag order m will use, indicate the length and lag item of the lag period to be considered
Number.Lag order is bigger, and the lag period of consideration is longer.In one embodiment, this is manually set according to actual needs
Lag order.In another embodiment, according to some algorithms, preferably lag order is determined.In the VAR model of subsequent builds
In, lag order m can be used in the same manner.Therefore, in the building detailed description lag order of subsequent combination VAR model
It determines.
In one embodiment, on the basis of ensuring that each variable and the liquidity scale have causal influence relationship, base
Column vector under each historical period obtained in step 25 constructs vector auto regression VAR model in step 27, determines
The model parameter of VAR model.
Vector Autoression Models abbreviation VAR model is proposed that it is each interior raw for 1980 by Christoffer simms
Variable constructs model as the function of endogenous variable lagged values all in system, so that having evaded needs in building model method
It will modeling problem to endogenous variable each in system about all endogenous variable lagged values.VAR model is usually used to prediction
Dynamic effects of the Time-series System and Disturbance to interact to system.
The mathematic(al) representation of VAR model is：
Yt=Φ1Yt-1+Φ2Yt-2+…+ΦmYt-m+εt (3)
Wherein, YtIt is k dimension endogenous variable, indicates the column vector of t moment, Yt-iIndicate the column vector at t-i moment, m is stagnant
Order afterwards, Φ1,Φ2..., ΦmIt is k*k dimension parameter matrix to be estimated, εtDisturbance column vector is tieed up for k, or is disturbance term.
During constructing Vector Autoression Models VAR, generally, first have to determine above-mentioned lag order m.Pass through
The expression formula of model above can be seen that the length that lag order m reflects the lag period of model consideration.Therefore, a side
Face, it is expected that lag order is sufficiently large, so that the model constructed can more completely reflect the dynamic that object changes over time
Feature；On the other hand, lag order is bigger, and the parameter for needing to estimate is more, and the freedom degree of model will be reduced.Therefore, right
In specific application scenarios, preferably lag order can be determined, thus the considerations of preferably balancing these two aspects.
In one embodiment, it is examined using likelihood ratio LR and determines lag order m.In LR inspection, from maximum lag
Order starts to examine.Null hypothesis H0：When lag order is j, parameter matrix ΦjElement be 0；Alternative hypothesis H1：Parameter
Matrix ΦjIn at least one element significantly not be 0.Then LR statistic is calculated, if the statistic is greater than critical value,
Refuse null hypothesis；Otherwise, receive null hypothesis, then reduce by a lag order and test, until refusing null hypothesis.
In one embodiment, according to red pond information criterion also known as minimal information amount AIC information criterion, lag rank is determined
Number.According to AIC information criterion, defining AIC function is：AIC=-2ln (maximum likelihood function value in model)+2 is (in model not
Know number of parameters).So that AIC function is reached the model of minimum value is considered as optimal models, correspondingly, reaches AIC function most
The lag order of small value is considered as optimal lag order.
Similar, lag order can also be determined according to Schwarz information criterion SC.In SC information criterion, SC function
It is defined as：SC=-2ln (maximum likelihood function value in model)+ln (unknown parameter number in model).SC function is reached
The lag order of minimum value is determined as optimal lag order.
Those skilled in the art can also determine above-mentioned lag order using other modes.
In addition, as previously mentioned, before or while constructing VAR model carry out Granger CaFpngerusality test in the case where, Glan
Also lag order can be used during outstanding Causality Test.Generally, by the lag order in Granger CaFpngerusality test be arranged with
It is identical in VAR model.Therefore, in one embodiment, it can determine a preferably lag order, while be used for Glan
Both outstanding Causality Test and VAR model.If when selecting variable (the first variable/the second variable), by other means
The variable and the liquidity scale for ensuring selection have causal influence relationship, then the step of Granger CaFpngerusality test can also be saved
Suddenly, above-mentioned lag order is determined when constructing VAR model.
On the basis of lag order has been determined, so that it may by the column under each historical period obtained in step 25 to
Amount, be input to more than formula (3) in, so that it is determined that other model parameters of VAR model, including parameter matrix Φ1,
Φ2..., ΦmWith disturbance term εt。
For example, formula (3) can be written as so that lag order m is 3 as an example：
Yt=Φ1Yt-1+Φ2Yt-2+Φ3Yt-3+εt (4)
Column vector under each historical period that step 25 obtains can be expressed as Yt=(at,bt,ct), wherein a is first
Variable, b are the second variable, and c is the liquidity scale.By the time window t, available multiple groups Y that change historical periodt,Yt-1,
Yt-2,Yt-3, these data are substituted into VAR model formation respectively, so that it may by calculating and training, determine model parameter
Φ1,Φ2, Φ3And εt。
In this way, constructing VAR model and determining the model parameter of VAR model, these models can will be then contained
The VAR model of parameter is as predicting the prediction model of the liquidity scale.
When being predicted using the VAR model constructed above, obtain first the column of recent continuous multiple periods to
Amount, is inputted VAR model formation (3), since parameter matrix and disturbance term are equal it has been determined that can be direct by formula (3)
The column vector of subsequent period is calculated.The first variable, the second variable and flowing in each column vector comprising the corresponding period
Property index.Therefore, the liquidity scale value can be extracted from the column vector of the subsequent period of output, that as predicts is next
The liquidity scale of period.
Still by taking lag order m=3 as an example, formula (4) can be rewritten as：
Yt+1=Φ1Yt+Φ2Yt-1+Φ3Yt-2+ε
As the column vector Y that each variable of input today (t) is constitutedt, column vector Y that yesterday (t-1), each variable was constitutedt-1, with
And the column vector Y that the day before yesterday (t-2) each variable is constitutedt-2, so that it may it is calculated according to above formula, is predicted in other words, tomorrow (t+1) is each
The column vector Y that vector is constitutedt+1, the liquidity scale c is extracted from the column vector of the prediction, the flowing for the tomorrow as predicted
Property index.It is, of course, also possible to the column vector based on existing yesterday, today, and the column vector of tomorrow estimated, further
Estimate posteriori column vector.It is appreciated that prediction accuracy may reduce as predicted time increases.But for short
The prediction of the liquidity scale of phase, the VAR model so constructed is highly effective.
On the basis of constructing VAR prediction model, in one embodiment, also the VAR model of above-mentioned building is carried out
Pulse response analysis determines the first variable/second variable to the pulse effects of the liquidity scale, to be conducive to building more
Accurate Short-term Forecasting Model.
Pulse response analysis is the dynamic effects analyzed when VAR model is impacted by certain to system.More specifically,
It is described on the stochastic error of some endogenous variable after the impact of one standard deviation size of application, to all interior lifes
Working as variable influences caused by time value and future value.If Granger CaFpngerusality test is qualitatively to analyze a variable to be
It is no to have an impact to another variable, then pulse response analysis then can quantitatively analyze the impact of a variable to another
Variable effect.In general, what pulse response analysis considered is short-term effect between variable.
Referring still to the mathematic(al) representation formula (3) of VAR model, wherein the disturbance term is also a k there are disturbance term
Dimensional vector, also known as impact vector, they can be contemporaneous correlation between each other, but it is unrelated with the lagged value of oneself and
It is unrelated with the variable on the right of equation.
In order to analyze influence of the pulse shock to its dependent variable of a variable, impulse response function is defined：
Cij indicate, other error terms it is in office when the phase is all constant under conditions of, when the corresponding error term of j-th of variable
After the t phase is by the impact of a unit, i-th of endogenous variable is influenced caused by the t+q phase.It is rung by above pulse
Function, influence of the impact of situational variables j to variable i in the t+q phase are answered, to more accurately assess the short-term shadow between variable
It rings.
In conjunction with the VAR model of the prediction the liquidity scale constructed above, some first variable or the second variable can be directed to
Establish above-mentioned impulse response function, analyze the impacts of the particular variables to the liquidity scale scheduled duration (t+q phase) shadow
It rings.In one embodiment, based on the impulse response function as above established, the particular variables are established to mobility Index Influence
Short-term forecast submodel.
In particular, the first variable is change relevant to Central Bank's currency operations scale in this specification one embodiment
Amount.By pulse response analysis, such as can using the liquidity scale as the above variable i using the first variable as the above variable j
More accurately to determine, Central Bank's monetary policy impacts the pulsing effect effect to the liquidity scale.
As above, VAR model is constructed for predicting the liquidity scale of commutative resource.In particular, by the prediction model
When for predicting Inter-Bank Market the liquidity scale, prediction and forecast can be provided for the liquidity scale, rather than as the past one
Sample can only carry out " summary " using these indexs.The Predict of the liquidity scale and forecast can be Central Bank money market
The trade decision and funding arrangement of operation and financial institution provide perspective guidance, field of plugging a gap.This also more meets stream
That is, the core and future developing trend of dynamic property risk management, identify potential liquidity risk perspectively, to realize early stage
Early warning and anticipation.
According to the embodiment of another aspect, a kind of device for constructing prediction model is also provided, the prediction model is for predicting
The liquidity scale of commutative resource.Fig. 3 shows the schematic block diagram of the model construction device according to one embodiment.Such as Fig. 3
Shown, model construction device 300 includes：Retrieval unit 31 is configured to obtain the first Variables Sequence, the second Variables Sequence
And index series, first Variables Sequence and the second Variables Sequence are respectively, in the multiple history being sequentially arranged
Period respectively to the first variable and the second variable carry out sampling formation Variables Sequence, wherein first variable and it is described can
The scale of operation of exchange resource is associated, and second variable is associated with the return rate of the commutative resource；The finger
Mark sequence includes the corresponding the liquidity scale of the multiple historical period；Stationarity detection unit 33 is configured to detect
The stationarity of first Variables Sequence, the second Variables Sequence and index series, to obtain stable first Variables Sequence, second
Variables Sequence and index series；Vector determination unit 35 is configured to stable first Variables Sequence, the second Variables Sequence
And index series, determine the corresponding column vector of each historical period, wherein the corresponding column vector Yi of the i-th historical period includes, it should
The liquidity scale of first variable of the i-th historical period, the second variable and i-th historical period；And parameter determines list
Member 37 is configured to the corresponding column vector input vector autoregression VAR model of each historical period determining VAR model
Model parameter, thus using the VAR model comprising above-mentioned model parameter as the prediction model.
According to a kind of embodiment, the first variable includes one or more of following variable：Reverse back purchase, medium term lending
Convenient MLF, ultra-short term reverse back purchase SLO, and the convenient SLF of standing debt-credit mortgages supplementary loan PSL；
Second variable includes one or more groups of in following variable：Repo rate and trading volume；Inter-bank rate and friendship
Yi Liang；Interest rate swap interest rate and trading volume；Yield of public debt and trading volume.
In one embodiment, stationarity detection unit 33 is configured to：Using unit root test, to detect the first variable
The sequence of unit root will be present as unstable sequence in the stationarity of sequence, the second Variables Sequence and index series.
According to one embodiment, stationarity detection unit 33 is additionally configured to：Detecting the first Variables Sequence, the second change
It measures in sequence and index series in the jiggly situation of any sequence, calculus of differences is carried out to jiggly sequence, and again
It tests, until getting stable first Variables Sequence, the second Variables Sequence and index series.
In one embodiment, device 300 further includes cause and effect detection unit 34, is configured to：To the first Variables Sequence/the
Two Variables Sequences carry out the Granger causality test with index series, to examine the first variable/second variable and mobility
Causal influence relationship between index.
Specifically, in one embodiment, by the inspection of cause and effect detection unit 34, it is ensured that the first variable and second becomes
At least one variable and the liquidity scale in amount have Granger two-way causal relationship.
According to a kind of embodiment, 34 concrete configuration of cause and effect detection unit is：Determine lag order m；Based on lag rank
Number, carries out the first recurrence processing to the lag item of the liquidity scale, returns processing result to the first variable/the second based on first
The lag item of variable carries out the second recurrence processing, obtains statistic；The first change is examined using F inspection based on the statistic
Measuring the/the second variable, whether there are Granger causalities with the liquidity scale.
According to a kind of embodiment, the VAR model of building is used for, and for lag order m, is based on continuous m historical period
Column vector, corresponding m parameter matrix and disturbance term, determine that next after the continuous m historical period goes through
History period corresponding column vector.Correspondingly, parameter determination unit 37 is configured to：Determine lag order m；It each is gone through based on described
History period corresponding column vector determines the m parameter matrix and the disturbance term.
In one embodiment, cause and effect detection unit 34 or parameter determination unit 37 determine lag in the following manner
Order m：It is examined using likelihood ratio LR and determines lag order m；Alternatively, determining lag order according to akaike information criterion AIC
m；Alternatively, determining lag order m according to Schwarz information criterion SC.
In one embodiment, device 300 further includes pulse analysis unit 38, is configured to carry out for the VAR model
Pulse response analysis determines the first variable/second variable to the pulse effects of the liquidity scale.
Above device 300 can by it is any have calculating, the device of processing capacity, equipment, platform, device clusters Lai
It realizes.In one embodiment, which, which can also embody, is independent financial product, pre- to provide the liquidity scale
The service of survey, or it is presented as product module, it is incorporated into financial analysis Evaluation Platform.
By the method and apparatus of above embodiments, prediction model can be targetedly constructed, to be the liquidity scale
Perspective forecast analysis is provided, the blank of warning aspect in liquidity risk management is filled up.
According to the embodiment of another aspect, a kind of computer readable storage medium is also provided, is stored thereon with computer journey
Sequence enables computer execute and combines method described in Fig. 2 when the computer program executes in a computer.
According to the embodiment of another further aspect, a kind of calculating equipment, including memory and processor, the storage are also provided
It is stored with executable code in device and realizes method described in conjunction with Figure 2 when the processor executes the executable code.
Those skilled in the art are it will be appreciated that in said one or multiple examples, function described in the invention
It can be realized with hardware, software, firmware or their any combination.It when implemented in software, can be by these functions
Storage in computer-readable medium or as on computer-readable medium one or more instructions or code passed
It is defeated.
Above-described specific embodiment has carried out further the purpose of the present invention, technical scheme and beneficial effects
It is described in detail, it should be understood that being not used to limit this hair the foregoing is merely a specific embodiment of the invention
Bright protection scope, all any modification, equivalent substitution, improvement and etc. on the basis of technical solution of the present invention, done,
It should all include within protection scope of the present invention.
Claims (22)
1. a kind of method for constructing prediction model, the prediction model is used to predict the liquidity scale of commutative resource, described
Method includes：
Obtain the first Variables Sequence, the second Variables Sequence and index series, first Variables Sequence and the second Variables Sequence
Respectively, the change of sampling formation is carried out to the first variable and the second variable respectively in the multiple historical periods being sequentially arranged
Measure sequence, wherein first variable is associated with the scale of operation of the commutative resource, second variable with it is described can
The return rate of exchange resource is associated；The index series includes, the corresponding the liquidity scale of the multiple historical period；
The stationarity of the first Variables Sequence, the second Variables Sequence and index series is detected, to obtain stable first variable sequence
Column, the second Variables Sequence and index series；
Based on stable first Variables Sequence, the second Variables Sequence and index series, determine each historical period it is corresponding arrange to
Amount, wherein the corresponding column vector Yi of the i-th historical period includes, the first variable, the second variable of i-th historical period, and should
The liquidity scale of i-th historical period；
By the corresponding column vector input vector autoregression VAR model of each historical period, the model ginseng of VAR model is determined
Number, thus using the VAR model comprising above-mentioned model parameter as the prediction model.
2. according to the method described in claim 1, wherein,
First variable includes one or more of following variable：Reverse back purchase, the convenient MLF of medium term lending, ultra-short term reverse back
SLO is purchased, the convenient SLF of standing debt-credit mortgages supplementary loan PSL；
Second variable includes one or more groups of in following variable：Repo rate and trading volume；Inter-bank rate and friendship
Yi Liang；Interest rate swap interest rate and trading volume；Yield of public debt and trading volume.
3. according to the method described in claim 1, wherein, detection the first Variables Sequence, the second Variables Sequence and the index sequence
The stationarity of column includes：Using unit root test, to detect the steady of the first Variables Sequence, the second Variables Sequence and index series
Property, the sequence of unit root will be present as unstable sequence.
4. according to the method described in claim 1, wherein, detecting the first Variables Sequence, the second Variables Sequence and index series
Stationarity, to obtain stable first Variables Sequence, the second Variables Sequence and index series and include：
The first Variables Sequence is being detected, in the second Variables Sequence and index series in the jiggly situation of any sequence, to this
Jiggly sequence carries out calculus of differences, and tests again, until getting stable first Variables Sequence, the second variable
Sequence and index series.
5. according to the method described in claim 1, further including：
Granger causality test with index series is carried out to the first Variables Sequence/second Variables Sequence, to examine first
The causal influence relationship of variable/between the second variable and the liquidity scale.
6. according to the method described in claim 5, wherein first variable and at least one variable and institute in the second variable
The liquidity scale is stated with Granger two-way causal relationship.
7. according to the method described in claim 5, wherein being carried out and index series to the first Variables Sequence/second Variables Sequence
Granger causality test includes：
Determine lag order m；
Based on lag order, the first recurrence processing is carried out to the lag item of the liquidity scale, returns processing result pair based on first
The bivariate lag item of first variable/the carries out the second recurrence processing, obtains statistic；
Based on the statistic examined using F inspection the first variable/second variable whether with the liquidity scale there are Grangers
Causality.
8., for lag order m, being gone through based on continuous m according to the method described in claim 1, wherein the VAR model is used for
The column vector of history period, corresponding m parameter matrix and disturbance term, under determining after the continuous m historical period
The corresponding column vector of one historical period；The model parameter of the determining VAR model includes：
Determine lag order m；
Based on the corresponding column vector of each historical period, the m parameter matrix and the disturbance term are determined.
9. method according to claim 7 or 8, wherein the determining lag order m includes：
It is examined using likelihood ratio LR and determines lag order m；Or
According to akaike information criterion AIC, lag order m is determined；Or
According to Schwarz information criterion SC, lag order m is determined.
10. carry out pulse response analysis for the VAR model according to the method described in claim 1, further include, the is determined
Pulse effects of one variable/second variable to the liquidity scale.
11. a kind of device for constructing prediction model, the prediction model is used to predict the liquidity scale of commutative resource, described
Device includes：
Retrieval unit is configured to obtain the first Variables Sequence, the second Variables Sequence and index series, first variable
Sequence and the second Variables Sequence are respectively to become respectively to the first variable and second in the multiple historical periods being sequentially arranged
Amount carries out the Variables Sequence of sampling formation, wherein first variable is associated with the scale of operation of the commutative resource, institute
It is associated with the return rate of the commutative resource to state the second variable；The index series includes the multiple historical period point
Not corresponding the liquidity scale；
Stationarity detection unit is configured to the stationarity of the first Variables Sequence of detection, the second Variables Sequence and index series, thus
Obtain stable first Variables Sequence, the second Variables Sequence and index series；
Vector determination unit is configured to stable first Variables Sequence, the second Variables Sequence and index series, determines each
The corresponding column vector of historical period, wherein the corresponding column vector Yi of the i-th historical period includes, the first of i-th historical period becomes
The liquidity scale of amount, the second variable and i-th historical period；
Parameter determination unit is configured to by the corresponding column vector input vector autoregression VAR model of each historical period, really
The model parameter of VAR model is determined, thus using the VAR model comprising above-mentioned model parameter as the prediction model.
12. device according to claim 11, wherein
First variable includes one or more of following variable：Reverse back purchase, the convenient MLF of medium term lending, ultra-short term reverse back
SLO is purchased, the convenient SLF of standing debt-credit mortgages supplementary loan PSL；
Second variable includes one or more groups of in following variable：Repo rate and trading volume；Inter-bank rate and friendship
Yi Liang；Interest rate swap interest rate and trading volume；Yield of public debt and trading volume.
13. device according to claim 11, wherein the stationarity detection unit is configured to：Using unit root test,
It detects the stationarity of the first Variables Sequence, the second Variables Sequence and index series, the sequence of unit root will be present as uneven
Steady sequence.
14. device according to claim 11, wherein the stationarity detection unit is configured to：
The first Variables Sequence is being detected, in the second Variables Sequence and index series in the jiggly situation of any sequence, to this
Jiggly sequence carries out calculus of differences, and tests again, until getting stable first Variables Sequence, the second variable
Sequence and index series.
15. device according to claim 11 further includes cause and effect detection unit, is configured to：
Granger causality test with index series is carried out to the first Variables Sequence/second Variables Sequence, to examine first
The causal influence relationship of variable/between the second variable and the liquidity scale.
16. device according to claim 15, wherein at least one variable in first variable and the second variable with
The liquidity scale has Granger two-way causal relationship.
17. device according to claim 15, wherein the cause and effect detection unit is configured to：
Determine lag order m；
Based on lag order, the first recurrence processing is carried out to the lag item of the liquidity scale, returns processing result pair based on first
The bivariate lag item of first variable/the carries out the second recurrence processing, obtains statistic；
Based on the statistic examined using F inspection the first variable/second variable whether with the liquidity scale there are Grangers
Causality.
18. device according to claim 11, wherein the VAR model is used for, for lag order m, based on continuous m
The column vector of historical period, corresponding m parameter matrix and disturbance term determine after the continuous m historical period
The corresponding column vector of next historical period；The parameter determination unit is configured to：
Determine lag order m；
Based on the corresponding column vector of each historical period, the m parameter matrix and the disturbance term are determined.
19. device described in 7 or 18 according to claim 1, wherein the cause and effect detection unit or the parameter determination unit are logical
It crosses following manner and determines lag order m：
It is examined using likelihood ratio LR and determines lag order m；Or
According to akaike information criterion AIC, lag order m is determined；Or
According to Schwarz information criterion SC, lag order m is determined.
20. being configured to carry out arteries and veins for the VAR model according to the method described in claim 1, further including pulse analysis unit
Response analysis is rushed, determines the first variable/second variable to the pulse effects of the liquidity scale.
21. a kind of computer readable storage medium, is stored thereon with computer program, when the computer program in a computer
When execution, computer perform claim is enabled to require the method for any one of 1-10.
22. a kind of calculating equipment, including memory and processor, which is characterized in that be stored with executable generation in the memory
Code realizes method of any of claims 1-10 when the processor executes the executable code.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810707589.7A CN108921688A (en) | 2018-07-02 | 2018-07-02 | Construct the method and device of prediction model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810707589.7A CN108921688A (en) | 2018-07-02 | 2018-07-02 | Construct the method and device of prediction model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108921688A true CN108921688A (en) | 2018-11-30 |
Family
ID=64423482
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810707589.7A Pending CN108921688A (en) | 2018-07-02 | 2018-07-02 | Construct the method and device of prediction model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108921688A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110363661A (en) * | 2019-08-05 | 2019-10-22 | 中国工商银行股份有限公司 | Bank liquidity prediction technique and device |
| CN110365708A (en) * | 2019-08-05 | 2019-10-22 | 山东浪潮人工智能研究院有限公司 | A kind of exchange data method for detecting abnormality based on Vector Autoression Models |
| CN111163092A (en) * | 2019-12-30 | 2020-05-15 | 深信服科技股份有限公司 | Flow abnormity detection method, device, equipment and storage medium |
| CN111259537A (en) * | 2020-01-14 | 2020-06-09 | 中交第二公路勘察设计研究院有限公司 | Road surface performance prediction method based on VAR (variable-offset-ratio) multivariate time sequence |
| CN112149880A (en) * | 2020-09-02 | 2020-12-29 | 中移（杭州）信息技术有限公司 | User scale prediction method, device, electronic equipment and storage medium |
| CN112925271A (en) * | 2019-12-06 | 2021-06-08 | 阿里巴巴集团控股有限公司 | Data processing method, automatic control method and device |
| CN113379436A (en) * | 2020-03-09 | 2021-09-10 | 阿里巴巴集团控股有限公司 | Information processing method, device, computing equipment and medium |
| CN113554449A (en) * | 2020-04-23 | 2021-10-26 | 阿里巴巴集团控股有限公司 | Commodity variable prediction method, commodity variable prediction device, and computer-readable medium |
-
2018
- 2018-07-02 CN CN201810707589.7A patent/CN108921688A/en active Pending
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110363661A (en) * | 2019-08-05 | 2019-10-22 | 中国工商银行股份有限公司 | Bank liquidity prediction technique and device |
| CN110365708A (en) * | 2019-08-05 | 2019-10-22 | 山东浪潮人工智能研究院有限公司 | A kind of exchange data method for detecting abnormality based on Vector Autoression Models |
| CN110365708B (en) * | 2019-08-05 | 2021-12-07 | 山东浪潮科学研究院有限公司 | Switchboard data anomaly detection method based on vector autoregressive model |
| CN112925271A (en) * | 2019-12-06 | 2021-06-08 | 阿里巴巴集团控股有限公司 | Data processing method, automatic control method and device |
| CN111163092A (en) * | 2019-12-30 | 2020-05-15 | 深信服科技股份有限公司 | Flow abnormity detection method, device, equipment and storage medium |
| CN111259537A (en) * | 2020-01-14 | 2020-06-09 | 中交第二公路勘察设计研究院有限公司 | Road surface performance prediction method based on VAR (variable-offset-ratio) multivariate time sequence |
| CN111259537B (en) * | 2020-01-14 | 2023-03-31 | 中交第二公路勘察设计研究院有限公司 | Road surface performance prediction method based on VAR (variable-value response) multivariate time sequence |
| CN113379436A (en) * | 2020-03-09 | 2021-09-10 | 阿里巴巴集团控股有限公司 | Information processing method, device, computing equipment and medium |
| CN113554449A (en) * | 2020-04-23 | 2021-10-26 | 阿里巴巴集团控股有限公司 | Commodity variable prediction method, commodity variable prediction device, and computer-readable medium |
| CN112149880A (en) * | 2020-09-02 | 2020-12-29 | 中移（杭州）信息技术有限公司 | User scale prediction method, device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108921688A (en) | Construct the method and device of prediction model | |
| Aue et al. | Sequential testing for the stability of high-frequency portfolio betas | |
| Menkhoff et al. | Heterogeneity in exchange rate expectations: Evidence on the chartist–fundamentalist approach | |
| Kouwenberg et al. | Model uncertainty and exchange rate forecasting | |
| Mazumder | The stability of the Phillips curve in India: Does the Lucas critique apply? | |
| Ma et al. | Macroeconomic attention and stock market return predictability | |
| Narayan et al. | The financial econometrics of price discovery and predictability | |
| Gelper et al. | Multivariate out-of-sample tests for Granger causality | |
| TW201514894A (en) | System and method for rating and selecting models | |
| OZKAN et al. | Agricultural value added and economic growthin the European Union accession process | |
| Delle Monache et al. | Adaptive state space models with applications to the business cycle and financial stress | |
| Arestis et al. | House price overvaluation in Hong Kong: Identifying the market fundamentals and understanding the “Bubble” | |
| Ulm et al. | Do interest rate differentials drive the volatility of exchange rates? Evidence from an extended stochastic volatility model | |
| CN114493208A (en) | Method and device for evaluating engineering project full life cycle, electronic equipment and medium | |
| CN109598607A (en) | Method, apparatus and storage medium based on artificial intelligence monitoring self learning model | |
| Chauvet et al. | What does realized volatility tell us about macroeconomic fluctuations? | |
| Guarin et al. | Recovering default risk from CDS spreads with a nonlinear filter | |
| Saidane | A Monte-Carlo-based Latent Factor Modeling Approach with Time-Varying Volatility for Value-at-Risk Estimation: Case of the Tunisian Foreign Exchange Market: Case of the Tunisian Foreign Exchange Market | |
| Reyes-Heroles et al. | Interest rate volatility and sudden stops: An empirical investigation | |
| Ranjan et al. | The Dynamics of Money Velocity, External Sectors and Electronic Transactions in India: Connecting dots using Empirical Approach. | |
| Bentes et al. | On the predictability of realized volatility using feasible GLS | |
| Mwamba et al. | Implied volatility of foreign exchange options: A leading indicator for currency crisis identification | |
| Huschens et al. | Country default probabilities: assessing and backtesting | |
| Pavel et al. | Decision Making Process of Stock Trading Implementing DRQN And ARIMA | |
| Berggren | The estimation of factors in FAVAR models |
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 | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20200923 Address after: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant after: Advanced innovation technology Co.,Ltd. Address before: A four-storey 847 mailbox in Grand Cayman Capital Building, British Cayman Islands Applicant before: Alibaba Group Holding Ltd. Effective date of registration: 20200923 Address after: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant after: Innovative advanced technology Co.,Ltd. Address before: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant before: Advanced innovation technology Co.,Ltd. |
|
| TA01 | Transfer of patent application right | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181130 |
|
| RJ01 | Rejection of invention patent application after publication |