CN105631744A - Method and apparatus for visually displaying high-frequency financial time sequence correlation - Google Patents

Abstract

The invention discloses a method for visually displaying high-frequency financial time sequence correlation. The method comprises the following steps: obtaining n groups of high-frequency financial data sequences, wherein n is greater than and equal to two; and processing the high-frequency financial data sequences to obtain n*n-dimensional matrix sequence; and processing the matrix sequence by use of a slide window mode and an investigation width of w to form a window including n*n*w items, combining any items among the n*n*w items to form item set sequences, recording the frequency of the item set sequences occurring in the window in the slide window mode, and screening a frequent sequence formed by an item set sequence whose support number is greater than and equal to a preset support number threshold from the item set sequences, wherein the frequency is the support number of the item set sequences. The invention provides a device for visually displaying high-frequency financial time sequence correlation. According to the technical demand for financial time sequence analysis, correlation between multiple financial time sequence data can be visually and rapidly determined and the existence mode of a correlation status is discovered.

Description

The method of visual presentation high-frequency transaction data dependency and device

Technical field

The present invention relates to finance statistics field, in particular to method and the device of a kind of visual presentation high-frequency transaction data dependency.

Background technology

Method general in existing market adopts display curve to carry out indirect labor to judge the correlationship between the assets such as such as stock; General employing three kinds of methods: (1) Drawing Directly method: this method is used by the stock exchange trading system of great majority; Raw data or its simple transformation result is used directly to draw figure; General conventional scatter diagram or graphic representation; By degree of correlation is determined in the manual observation of figure. (2) correlation curve: this method is generally use when assets are carried out Depth Study by specialty investment corporation; The discrete partial auto-correlation of this method by calculating between two time serieses, draws correlation curve between the two, thus visual reflection between the two conditions associated. (3) correlation matrix heat figure: use when this method is generally specialized company or the researcher dependency situation between the multiple sequence of Depth Study; The correlation matrix of the method by calculating between multiple sequence, and then this correlation matrix shows as " heat figure " realize visual, namely uses relatively dark colour rectangular block to represent bigger relation conefficient, represents less relation conefficient with compared with the rectangular block of light colour.

Above-mentioned three kinds of methods all can use single computer to run special software and realize, it is possible to use server provides the form of webpage to realize by internet, it is also possible to use mobile phone A PP to be displayed on mobile phone screen.

But above-mentioned three kinds of methods all also exist shortcoming: (1) method that directly raw data is drawn, it is necessary to rely on the degree of artificial judgment dependency, lacks accuracy, and do not have objective judging criterion. (2) method of relation conefficient curve, can only reflect that one to the dependency situation over time of data sequence, can not reflect the situation of dependency between multiple data. (3) method such as hot figure of correlation matrix, although the dependency state between multiple data sequence can be reflected, but it can not reflect dependency state situation over time, does not also reflect different associative modes and the measure of probability thereof. Meanwhile, the common shortcoming of above-mentioned three kinds of methods, be can not in real time, the dynamically pattern of correlation behavior between visual presentation financial time series and rule.

In finance data analysis field, it is necessary to the dependency between multiple financial asset is made analysis; Due to the complicacy of financial business, to the judgement of affairs in complex system, it is difficult to make decision-making according to certain parameter value, and often need to use visual mode, after visual for some calculation result displaying, then made decision-making by considering of artificial intelligence by professional.

Correlation analysis (correlationanalysis) whether there is certain dependence between research phenomenon, and the phenomenon specifically having dependence is inquired into its related direction and degree of correlation, it is a kind of statistical method of the correlationship between research stochastic variable. The research of dependency (correlation) between financial asset is one of the core work that finance data is analyzed, for risk management, hedge, derived product price and optimal portfolio selection etc. all significant.

Can determine by formula below for two data sequence x and y, its relation conefficient C:

$C_{x,y}=\frac{{\mathrm{\Σ}}_{i=1}^n(x_i-\stackrel{\‾}{x})(y_i-\stackrel{\‾}{y})}{\sqrt{{\mathrm{\Σ}}_{i=1}^n{(x_i-\stackrel{\‾}{x})}^2}\sqrt{{\mathrm{\Σ}}_{i=1}^n{(y_i-\stackrel{\‾}{y})}^2}}$

Size according to C value, it is possible to judge the degree of correlation between x and y two sequences. Such as,

| C | > 0.95: there is significance and be correlated with;

| C | >=0.8 is highly relevant;

< in 0.8, degree is relevant for 0.5��| C |;

0.3��| C | < 0.5 lower correlation;

| C | < 0.3 relation is extremely weak, it is believed that uncorrelated;

If C < 0, changing in the opposite direction of X and Y is described; This is referred to as " negative correlation ".

But, relation conefficient can only reflect the correlationship between two data sequences on the whole, and the data sequence of many non-linear process often shows different local correlations feature.

Summary of the invention

The present invention is intended at least solve one of problems of the prior art.

One of the technical problem to be solved in the present invention is to solve the problem that can intuitively and fast the dependency between multiple financial time series data be judged.

For solving the problems of the technologies described above, the present invention provides the device of a kind of visual presentation high-frequency transaction data dependency, and its step comprises:

Obtain n group high frequency finance data sequence, n >=2;

Described high frequency finance data sequence is processed, obtains the matrix sequence of n �� n dimension;

Adopting moving window mode and investigating width is that w processes described matrix sequence, formation comprises the window of n �� n �� w item, by the combination of any item in described n �� n �� w item, form item collection sequence, record the described item collection sequence number of times that window occurs in moving window mode, described number of times is the support number of described item collection sequence, filters out the item collection sequence formation Frequent episodes that support number is more than or equal to default support number threshold value in described item collection sequence.

Further, its step also comprises:

Described Frequent episodes is excavated, generates, according to excavating result, the heat figure temporally sequentially arranged.

Further, its step also comprises:

By JavaScript special efficacy, described heat figure is generated animation, by http protocol, described animation is exported.

Further, described high frequency finance data sequence is yield volatility.

Further, described matrix sequence is correlation matrix sequence.

Further, described high frequency finance data sequence is logarithm income sequence.

Further, described matrix sequence is transfer entropy matrix sequence.

For solving the problems of the technologies described above, the present invention also provides the device of a kind of visual presentation high-frequency transaction data dependency, comprising:

High-frequency data acquisition module, for obtaining n group high frequency finance data sequence, n >=2;

Matrix sequence constructing module, for described high frequency finance data sequence being processed, obtains the matrix sequence of n �� n dimension;

Frequent Pattern Mining module, by adopting moving window mode and investigation width to be that w processes described matrix sequence, formation comprises the window of n �� n �� w item, by the combination of any item in described n �� n �� w item, form item collection sequence, recording the described item collection sequence number of times that window occurs in moving window mode, described number of times is the support number of described item collection sequence, filters out the item collection sequence formation Frequent episodes that support number is more than or equal to default support number threshold value in described item collection sequence.

Further, also comprise:

Visual presentation module, for being excavated by described Frequent episodes, generates, according to excavating result, the heat figure temporally sequentially arranged.

Further, described animation also for described heat figure being generated animation by JavaScript special efficacy, by http protocol, is exported by described visual presentation module.

The useful effect of the present invention there are provided a kind of method of visual presentation high-frequency transaction data dependency, the present invention is from one pair of data sequence extension to multiple data sequence, by the correlation matrix between all data sequences, form a matrix sequence changed in time, above-mentioned matrix sequence carries out sequential mode mining, and uses method for visualizing to display excavation result; Therefore the technical need that the present invention can analyze according to financial time series, intuitively and fast to the dependency between multiple financial time series data judges, it has been found that its dependency situation there is pattern, and the Changing Pattern of this kind of pattern over time. Present invention also offers the device of a kind of visual presentation high-frequency transaction data dependency simultaneously.

Accompanying drawing explanation

Fig. 1 show the schema of the method for a kind of visual presentation high-frequency transaction data dependency of the embodiment of the present invention.

Fig. 2 show the composition schematic diagram of the device of a kind of visual presentation high-frequency transaction data dependency of the embodiment of the present invention.

Embodiment

The present invention hereafter will be described in conjunction with specific embodiments in detail. It is noted that the combination of the technology feature described in following embodiment or technology feature should not be considered as isolated, they can mutually be combined thus be reached better technique effect.

The present invention provides a kind of method of visual presentation high-frequency transaction data dependency, and as shown in Figure 1, its step comprises:

S1: obtain n group high frequency finance data sequence, n >=2; Described high frequency finance data sequence is yield volatility or logarithm income sequence;

S2: described high frequency finance data sequence processed, obtains the matrix sequence of n �� n dimension; Wherein, obtain correlation matrix sequence by described yield volatility, obtain transfer entropy matrix sequence by logarithm income sequence;

S3: adopting moving window mode and investigating width is that w processes described matrix sequence, formation comprises the window of n �� n �� w item, by the combination of any item in described n �� n �� w item, form item collection sequence, record the described item collection sequence number of times that window occurs in moving window mode, described number of times is the support number of described item collection sequence, filters out the item collection sequence formation Frequent episodes that support number is more than or equal to default support number threshold value in described item collection sequence;

S4: excavated by described Frequent episodes, generates, according to excavating result, the heat figure temporally sequentially arranged;

S5: by JavaScript special efficacy, generates animation by described heat figure, by http protocol, is exported by described animation.

As shown in Figure 2, the present invention also provides the device of a kind of visual presentation high-frequency transaction data dependency, comprising:

High-frequency data acquisition module 100, for obtaining n group high frequency finance data sequence, n >=2;

Matrix sequence constructing module 200, for described high frequency finance data sequence being processed, obtains the matrix sequence of n �� n dimension;

Frequent Pattern Mining module 300, by adopting moving window mode and investigation width to be that w processes described matrix sequence, formation comprises the window of n �� n �� w item, by the combination of any item in described n �� n �� w item, form item collection sequence, recording the described item collection sequence number of times that window occurs in moving window mode, described number of times is the support number of described item collection sequence, filters out the item collection sequence formation Frequent episodes that support number is more than or equal to default support number threshold value in described item collection sequence;

Visual presentation module 400, for being excavated by described Frequent episodes, generates, according to excavating result, the heat figure temporally sequentially arranged, by JavaScript special efficacy, described heat is schemed generation animation, by http protocol, exported by described animation.

Embodiment 1

For share price high-frequency data, it is contemplated that the observed data on certain time window, its logarithmic return is:

R_t=ln (P_t)-ln(P_t-1)

P in formula_tIt is the closing price of t day, earning rate R_tIt it is a time series.

Gather the time series x and y of two earning rates;

The Pearson correlation coefficient of x and y is:

$C_{x,y}=\frac{{\mathrm{\&Sigma;}}_{i=1}^n(x_i-\stackrel{\&OverBar;}{x})(y_i-\stackrel{\&OverBar;}{y})}{\sqrt{{\mathrm{\&Sigma;}}_{i=1}^n{(x_i-\stackrel{\&OverBar;}{x})}^2}\sqrt{{\mathrm{\&Sigma;}}_{i=1}^n{(y_i-\stackrel{\&OverBar;}{y})}^2}}$

In formula, i represents sampling; The relation conefficient composition correlation matrix of multiple assets accordingly, the correlation matrix on multiple time period forms correlation matrix sequence.

Such as, three income sequences 1,2,3 are at time t₁And t₂On, its relation conefficient forms two correlation matrixes:

${\left[\begin{array}{ccc}{c}_{1,1}& c_{1,2}& c_{1,3}\\ c_{2,1}& c_{2,2}& c_{2,3}\\ c_{3,1}& c_{3,2}& c_{3,3}\end{array}\right]}_{t1}$ With ${\left[\begin{array}{ccc}{c}_{1,1}& c_{1,2}& c_{1,3}\\ c_{2,1}& c_{2,2}& c_{2,3}\\ c_{3,1}& c_{3,2}& c_{3,3}\end{array}\right]}_{t2}$

Wherein, c in first matrix_1,3It is that sequence 1 and sequence 3 are at moment t₁Relation conefficient, and the c in the 2nd matrix_1,3It is that sequence 1 and sequence 3 are at moment t₂Relation conefficient. In the same way, it is to construct at moment t₃, t₄On correlation matrix; Therefore, between upper M sequence of N number of moment, the correlation matrix sequence that the correlation matrix of N number of M*M is formed can be constructed.

The correlation matrix sequence of acquisition is carried out denoising Processing according to Random Matrices Theory, obtains the correlation matrix sequence after de-noising.

Adopting moving window mode and investigating width is the correlation matrix sequence after w processes de-noising, formation comprises the window of n �� n �� w item, by the combination of any item in described n �� n �� w item, form item collection sequence, record the described item collection sequence number of times that window occurs in moving window mode, described number of times is the support number of described item collection sequence, filters out the item collection sequence formation Frequent episodes that support number is more than or equal to default support number threshold value in described item collection sequence;

Described Frequent episodes is excavated, generates, according to excavating result, the heat figure temporally sequentially arranged, by JavaScript special efficacy, described heat is schemed generation animation, by http protocol, described animation is exported.

Embodiment 2

Gather and a time period has 2 discrete logarithm income sequence x and y;

The state of x and y on moment n is designated as i respectively_nAnd j_n; Due to the short memory of logarithm income sequence, for the purpose of simple, it is assumed that this sequence is a rank Markov process, then the transfer entropy of x to y is:

${\mathrm{TE}}_{x\&RightArrow;y}=\underset{i_{n+1},i_n,j_n}{\mathrm{\&Sigma;}}p(i_{n+1},i_n,j_n){\log }_2\frac{p(i_{n+1},i_n,j_n)p\left(i_n\right)}{p(i_{n+1},i_n)p(i_n,j_n)}$

According to upper formula, transfer entropy matrix sequence can be constructed;

The transfer entropy matrix sequence of acquisition is carried out denoising Processing according to Random Matrices Theory, obtains the correlation matrix sequence after de-noising.

Adopting moving window mode and investigating width is the correlation matrix sequence after w processes de-noising, formation comprises the window of n �� n �� w item, by the combination of any item in described n �� n �� w item, form item collection sequence, record the described item collection sequence number of times that window occurs in moving window mode, described number of times is the support number of described item collection sequence, filters out the item collection sequence formation Frequent episodes that support number is more than or equal to default support number threshold value in described item collection sequence;

Described Frequent episodes is excavated, generates, according to excavating result, the heat figure temporally sequentially arranged, by JavaScript special efficacy, described heat is schemed generation animation, by http protocol, described animation is exported.

The present invention provides a kind of method of visual presentation high-frequency transaction data dependency, the present invention is from one pair of data sequence extension to multiple data sequence, by the correlation matrix between all data sequences, form a matrix sequence changed in time, above-mentioned matrix sequence carries out sequential mode mining, and uses method for visualizing to display excavation result; Therefore the technical need that the present invention can analyze according to financial time series, intuitively and fast to the dependency between multiple financial time series data judges, it has been found that its dependency situation there is pattern, and the Changing Pattern of this kind of pattern over time. Present invention also offers the device of a kind of visual presentation high-frequency transaction data dependency simultaneously.

Although having given some embodiments of the present invention herein, but having it will be understood by one skilled in the art that without departing from the spirit of the invention, it is possible to embodiment herein is changed. Above-described embodiment is exemplary, it should not using embodiment herein as the restriction of interest field of the present invention.

Claims (10)

1. the method for a visual presentation high-frequency transaction data dependency, it is characterised in that, its step comprises:

Obtain n group high frequency finance data sequence, n >=2;

Described high frequency finance data sequence is processed, obtains the matrix sequence of n �� n dimension;

Adopting moving window mode and investigating width is that w processes described matrix sequence, formation comprises the window of n �� n �� w item, by the combination of any item in described n �� n �� w item, form item collection sequence, record the described item collection sequence number of times that window occurs in moving window mode, described number of times is the support number of described item collection sequence, filters out the item collection sequence formation Frequent episodes that support number is more than or equal to default support number threshold value in described item collection sequence.

2. the method for visual presentation high-frequency transaction data dependency according to claim 1, it is characterised in that, its step also comprises:

Described Frequent episodes is excavated, generates, according to excavating result, the heat figure temporally sequentially arranged.

3. the method for visual presentation high-frequency transaction data dependency according to claim 2, it is characterised in that, its step also comprises:

By JavaScript special efficacy, described heat figure is generated animation, by http protocol, described animation is exported.

4. the method for visual presentation high-frequency transaction data dependency according to claim 1, it is characterised in that, described high frequency finance data sequence is yield volatility.

5. the method for visual presentation high-frequency transaction data dependency according to claim 4, it is characterised in that, described matrix sequence is correlation matrix sequence.

6. the method for visual presentation high-frequency transaction data dependency according to claim 1, it is characterised in that, described high frequency finance data sequence is logarithm income sequence.

7. the method for visual presentation high-frequency transaction data dependency according to claim 6, it is characterised in that, described matrix sequence is transfer entropy matrix sequence.

8. the device of a visual presentation high-frequency transaction data dependency, it is characterised in that, comprising:

High-frequency data acquisition module, for obtaining n group high frequency finance data sequence, n >=2;

Matrix sequence constructing module, for described high frequency finance data sequence being processed, obtains the matrix sequence of n �� n dimension;

Frequent Pattern Mining module, by adopting moving window mode and investigation width to be that w processes described matrix sequence, formation comprises the window of n �� n �� w item, by the combination of any item in described n �� n �� w item, form item collection sequence, recording the described item collection sequence number of times that window occurs in moving window mode, described number of times is the support number of described item collection sequence, filters out the item collection sequence formation Frequent episodes that support number is more than or equal to default support number threshold value in described item collection sequence.

9. the device of visual presentation high-frequency transaction data dependency according to claim 8, it is characterised in that, also comprise:

Visual presentation module, for being excavated by described Frequent episodes, generates, according to excavating result, the heat figure temporally sequentially arranged.

10. the device of visual presentation high-frequency transaction data dependency according to claim 9, it is characterized in that, described animation also for described heat figure being generated animation by JavaScript special efficacy, by http protocol, is exported by described visual presentation module.