CN103399867A - Self-adaptive adjusting method for linear combination prediction model weight - Google Patents

Abstract

The invention discloses a self-adaptive adjusting method for linear combination prediction model weight. The self-adaptive adjusting method is characterized in that the step 1, weight wi, m and the former k-1 calculating results of a prediction model fi except mth calculating result are related to the correlation coefficients of the former k actual results; the step 2, the weight wi, m and the mth calculating result of the prediction model fi are related to deviation of the actual results, and higher weight can be set according to the prediction model with the smaller deviation to the closest actual result; the step 3, the step 1 is calculated to obtain the correlation coefficients, the step 2 is calculated to obtain the deviation between the predicted results and the actual results, and accordingly the purpose of calculating self-adaptive weight is achieved. The self-adaptive adjusting method for the linear combination prediction model weight fully takes the deviation between the predicted results and the actual results and the correlation coefficients between the predicted results and the former k results so as to improve algorithm accuracy and algorithm reasonability.

Description

A kind of self-adapting regulation method of linear combination forecasting model weight

Technical field

The present invention relates to a kind ofly in tax index prediction,, according to predicting the outcome and actual result of linear combination forecasting model, dynamically change the method for the weight of each forecast model.

Background technology

In combination linear prediction, each forecast model role in prediction is not changeless, but is subjected to all kinds of extraneous factors to affect dynamic change.For example, in tax index prediction, each forecast model is subjected to the impact of macroeconomy situation, all kinds of economy and the tax policy.Due to the difference of all kinds of factors to the forecast model influence degree, the weight of each forecast model, also in dynamic change,, in order to guarantee the performance of linear combination forecasting model, need to dynamically be adjusted it adaptively.The applicant is new through looking into, and has retrieved one piece of relevant patent: weight adjusting module and weight regulating method [application publication number: CN1925543].In this patent, the inventor provides a kind of weight adjusting module and weight regulating method, is applicable to adjust the weight in the image-zooming technology.The described method of above-mentioned Patents invention is all for concrete application, and algorithm idea is adjusted Weight algorithm from the self-adaptation that this method adopts has essence different.

Summary of the invention

The purpose of this invention is to provide a kind of can be according to predicted value and actual result, the dynamic alignment combination forecasting

In i forecast model f _iWeights omega _i(i ∈ [1..n]) carries out the method for accommodation.

For reaching above purpose, the present invention takes following technical scheme to be achieved:

A kind of self-adapting regulation method of linear combination forecasting model weight, is characterized in that, in accordance with the following steps:

(1) determine initial weight: establishing the numerical value of actual result first is F ' ₀, each forecast model f _iPredicted value be F ' _i; , for each i ∈ [1..n], calculate

Each forecast model f _iInitial weight be Meet the model that prediction effect is good first and have higher initial weight, and

(2) dynamically adjust weight:

A, establish current total m time and predict the outcome and actual result, comprise except the m time before the F ' that predicts the outcome of k-1 time _{I, m-k+1}, F ' _{I, m-k+2}.., F ' _{I, m-1}, and the actual result F ' of front k time _M-k+1,

F′ _M-k+2.., F ' _m; Wherein, k is time window, defaultly gets 5; When m＜k, get k=m; Corresponding F ' _{I, m-k+1}, F ' _{I, m-k+2}.., F ' _{I, m-1}Weight be respectively w _{I, m-k+1}, w _{I, m-k+2}.., w _{I, m-1};

B, forecast model f _i, i ∈ [1..n], the result of calculation of front k-1 time and the related coefficient of front k actual result except the m time;

C, forecast model f _iThe deviation of the m time result of calculation and actual result,

D, according to the result of A, B, self-adaptation solves each forecast model f _iThe weights omega that (i ∈ [1..n]) adjusts for the m time _i,m,, for high with the actual result correlativity and less with actual result deviation forecast model, increase the value of weight; For with the low forecast model of actual result correlativity and the forecast model larger with the actual result deviation, reduce the value of weight.

In such scheme, in described step (2), the specific algorithm of B step is:

STEP1. calculate $F_m={\mathrm{\Σ}}_{i=1}^n{\mathrm{\ω}}_{i,m-1}{F}_{i,m}^{\′};$

Too frequently cause shake for fear of the weight adjustment, as F ' _m=F _m=0 or | F ' _m-F _m|/(| F ' _m|+| F _m|)≤5%, do not carry out the index weights adjustment, namely, for each i ∈ [1..n], make ω _i,m=ω _{I, m-1}, return to (ω _{1, m}, ω _{2, m}.., ω _i,m.., ω _n,m), algorithm withdraws from;

STEP2. according to forecast model f _iThe result of calculation of (i ∈ [1..n]) and the related coefficient of actual result, determine weight adjustment coefficient:

STEP2.1., for each i ∈ [1..n], adopt Pearson product-moment correlation coefficient (PPMCC) sequence of calculation F ' _{I, m-k+1}, F ' _{I, m-k+2}.., F ' _{I, m-1}With F ' _M-k+1, F ' _M-k+2.., F ' _m-1Related coefficient, obtain one group of correlation coefficient r _i, r _i∈ [1,1];

$r_i=\frac{{\mathrm{\Σ}}_{j=m-k+1}^{m-1}(F_{i,j}^{\′}-{\stackrel{\‾}{F}}_i^{\′})(F_j^{\′}-{\stackrel{\‾}{F}}^{\′})}{\sqrt{{\mathrm{\Σ}}_{j=m-k+1}^{m-1}{(F_{i,j}^{\′}-{\stackrel{\‾}{F}}_i^{\′})}^2{\mathrm{\Σ}}_{j=m-k+1}^{m-1}{(F_j^{\′}-{\stackrel{\‾}{F}}^{\′})}^2}}---\left(1\right)$

In formula (1), ${\stackrel{\‾}{F}}_i^{\′}={\mathrm{\Σ}}_{j=m-k+1}^{m-1}{F}_{i,j}^{\′}/(k-1),$ ${\stackrel{\‾}{F}}^{\′}={\mathrm{\Σ}}_{j=m-k+1}^{m-1}{F}_j^{\′}/(k-1);$

STEP2.2. to correlation coefficient r _i(i ∈ [1..n]) carries out following zero correction, obtains one group of weight and adjusts coefficient r ' _i, r ' _iShow more greatly forecast model f _iHigher with the actual result correlativity;

$r_i^{\′}=\left|r_i\right|-{\mathrm{\Σ}}_{i=1}^n\left|r_i\right|/n;---\left(2\right)$

In described step (2), the specific algorithm of C step is:

Calculate forecast model f _i(i ∈ [1..n]) the m time result of calculation F ' _i,mWith actual result F ' _mDeviation Δ F _i:

$\mathrm{\Δ}\stackrel{\‾}{F}=\frac{{\mathrm{\Σ}}_{i=1}^n|F_{i,m}^{\′}-F_m^{\′}|}{n};---\left(3\right)$

$\mathrm{\Δ}{F}_i=\mathrm{\Δ}\stackrel{\‾}{F}-|F_{i,m}^{\′}-F_m^{\′}|;---\left(4\right)$

Deviation Δ F _iLarger, show forecast model f _iThe m time forecasting accuracy high.

In described step (2), the specific algorithm of D step is:

STEP1. calculate forecast model f _iThe weight adjusted value Δ ω of (i ∈ [1..n]) _i,m:

STEP1.1. to correlation coefficient r ' _i(i ∈ [1..n]) and deviation Δ F _iSpan normalization:

$r_i^{\′}=\left|r_i\right|-{\mathrm{\Σ}}_{i=1}^n\left|r_i\right|/n(r_i^{\′}\∈[-\mathrm{1,1}\left]\right);---\left(5\right)$

$\mathrm{\Δ}{F}_i^{\′}=\frac{\mathrm{\Δ}{F}_i+\min \left(\mathrm{\Δ}{F}_i\right)}{\max \left(\mathrm{\Δ}{F}_i\right)-\min \left(\mathrm{\Δ}{F}_i\right)}(\mathrm{\Δ}{F}_i^{\′}\∈[-\mathrm{1,1}\left]\right);---\left(6\right)$

STEP1.2. calculate forecast model f _iThe weight adjusted value Δ ω of (i ∈ [1..n]) _i,m

$\mathrm{\&Delta;}{\mathrm{\&omega;}}_{i,m}={\mathrm{\&omega;}}_{i,m-1}\frac{(p_1{r}_i^{\&prime;}+{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="HDA00003472719200011.png"\; state="\{\{state\}\}">p</figure-callout>}}_2\mathrm{\&Delta;}{F}_i^{\&prime;})-{\mathrm{\&Sigma;}}_{i=1}^n(p_1{r}_i^{\&prime;}+{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="HDA00003472719200011.png"\; state="\{\{state\}\}">p</figure-callout>}}_2\mathrm{\&Delta;}{F}_i^{\&prime;})/n}{{\mathrm{\&Sigma;}}_{i=1}^n(p_1{r}_i^{\&prime;}+{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="HDA00003472719200011.png"\; state="\{\{state\}\}">p</figure-callout>}}_2\mathrm{\&Delta;}{F}_i^{\&prime;})}---\left(7\right)$

In formula (7), p ₁And p ₂Be respectively correlation coefficient r ' _i, deviation Δ F _iWeight, p wherein ₁+ p ₂=1, p under default setting ₁=0.4p ₂=0.6;

STEP2., for each i ∈ [1..n], calculate ω _i,m;

ω′ _i,m=ω _i,m-1+Δω _i,mi∈[1..n] (8)

${\mathrm{\&omega;}}_{i,m}=\frac{{\mathrm{\&omega;}}_{i,m}^{\&prime;}}{{\mathrm{\&Sigma;}}_{i=1}^n{\mathrm{\&omega;}}_{i,m}^{\&prime;}}---\left(9\right)$

Weight adjusted value Δ ω _i,mMeet

Return to (ω _{1, m}, ω _{2, m}.., ω _i,m.., ω _n,m).

Advantage of the present invention is to have utilized forecast model f _iThe m time outer before result of calculation and front k actual result related coefficient and the forecast model f of k-1 time _iThe m time result of calculation and the deviation of actual result increased accuracy rate and rationality that algorithm calculates.

Description of drawings

The present invention is described in further detail below in conjunction with the drawings and the specific embodiments.

Fig. 1 the inventive method process flow diagram.

Embodiment

The present invention adopts is according to the predicting the outcome and actual result of linear combination forecasting model, dynamically changes the method for the weight of each forecast model.

Research purpose:, according to predicting the outcome and actual result of linear combination forecasting model, dynamically change the weight of each tax index model.

Research background: in combination linear prediction, each forecast model role in prediction is not changeless, but is subjected to all kinds of extraneous factors to affect dynamic change.For example, in tax index prediction, each forecast model is subjected to the impact of macroeconomy situation, all kinds of economy and the tax policy.Due to the difference of all kinds of factors to the forecast model influence degree, the weight of each forecast model, also in dynamic change,, in order to guarantee the performance of linear combination forecasting model, need to dynamically be adjusted it adaptively.

As shown in Figure 1, a kind of self-adapting regulation method of linear combination forecasting model weight, comprise 6 steps, and its idiographic flow is:

(1) be F ' according to the numerical value of actual result first ₀With each forecast model f _iPredicted value be F ' _i,, for each i ∈ [1..n], calculate each forecast model f _iInitial weight w _i,0;

(2) too frequently cause shake for fear of the weight adjustment, if F ' _m=Fm=0 or | F ' _m-F _m|/(| F ' _m|+| F _m|)≤5%, do not carry out the index weights adjustment, make ω _i,m=ω _{I, m-1}, algorithm withdraws from, otherwise carries out (3).

(3) according to forecast model f _iThe result of calculation of (i ∈ [1..n]) and the related coefficient of actual result, determine weight adjustment coefficient:

(4) calculate forecast model f _i(i ∈ [1..n]) the m time result of calculation F ' _i,mWith actual result F ' _mRelative deviation Δ F _i:

(5) calculate forecast model f _iThe weight adjusted value Δ ω of (i ∈ [1..n]) _i,m;

(6), for each i ∈ [1..n], calculate ω _{I, m}=ω _{I, m-1}+ Δ ω _i,m; Return to (ω _{1, m}, ω _{2, m}.., ω _i,m.., ω _n,m), algorithm withdraws from.

According to following detailed step:

(1) determine initial weight: establishing the numerical value of actual result (actual value of tax index) first is F ' ₀, each forecast model f _iPredicted value be F ' _i; , for each i ∈ [1..n], calculate ${\mathrm{\&omega;}}_{i,0}^{\&prime;}=1-\frac{|F_0^{\&prime;}-F_i^{\&prime;}|}{|F_0^{\&prime;}+F_i^{\&prime;}|+1},$ Each forecast model f _iInitial weight be ${\mathrm{\&omega;}}_{i,0}=\frac{{\mathrm{\&omega;}}_{i,0}^{\&prime;}}{{\mathrm{\&Sigma;}}_{i=1}^n{\mathrm{\&omega;}}_{i,0}^{\&prime;}},$ Meet the model that prediction effect is good first and have higher initial weight, and

(2) dynamically adjust weight: establish current total m time and predict the outcome and actual result; For each i ∈ [1..n], according to forecast model f _iResult of calculation and front k the actual result related coefficient of k-1 time before except the m time, and forecast model f _iThe deviation of the m time result of calculation and actual result, solve each forecast model f _iThe weights omega that (i ∈ [1..n]) adjusts for the m time _i,m; For with high and less with the nearest actual result deviation forecast model of actual result correlativity in early stage, increase the value of weight; For with the low forecast model of actual result correlativity and the forecast model larger with nearest actual result deviation, reduce the value of weight.

Described dynamic adjustment weight is carried out according to following algorithm:

Input:

A) forecast model f _i(i ∈ [1..n]) k-1 time (k is time window, defaultly gets 5 before except the m time; When m＜k, get k=m) result of calculation be F ' _{I, m-k+1}, R ' _{I, m-k+2}.., F ' _{I, m-1};

B) corresponding F ' _{I, m-k+1}, F ' _{I, m-k+2}.., F ' _{I, m-1}Weight be respectively w _{I, m-k+1}, w _{I, m-k+2}.., w _{I, m-1};

C) the actual result F ' of front k time _M-k+1, F ' _M-k+2.., F ' _m;

Output: vector (ω _{1, m}, ω _{2, m}.., ω _i,m.., ω _n,m), ω _i,mFor forecast model f _iThe weight of the m time is adjusted result;

Step is:

STEP1. calculate $F_m={\mathrm{\&Sigma;}}_{i=1}^n{\mathrm{\&omega;}}_{i,m-1}{F}_{i,m}^{\&prime;};$

Too frequently cause shake for fear of the weight adjustment, as F ' _m=F _m=0 or | F ' _m-F _m|/(| F ' _m|+| F _m|)≤5%, do not carry out the index weights adjustment, namely, for each i ∈ [1..n], make ω _i,m=ω _{I, m-1}, return to (ω _{1, m}, ω _{2, m}.., ω _i,m.., ω _n,m), algorithm withdraws from;

STEP2. according to forecast model f _iThe result of calculation of (i ∈ [1..n]) and the related coefficient of actual result, determine weight adjustment coefficient:

STEP2.1., for each i ∈ [1..n], adopt Pearson product-moment correlation coefficient (PPMCC) sequence of calculation F ' _{I, m-k+1}, F ' _{I, m-k+2}.., F ' _{I, m-1}With F ' _M-k+1, F ' _M-k+2.., F ' _m-1Related coefficient, obtain one group of correlation coefficient r _i, r _i∈ [1,1];

$r_i=\frac{{\mathrm{\&Sigma;}}_{j=m-k+1}^{m-1}(F_{i,j}^{\&prime;}-{\stackrel{\&OverBar;}{F}}_i^{\&prime;})(F_j^{\&prime;}-{\stackrel{\&OverBar;}{F}}^{\&prime;})}{\sqrt{{\mathrm{\&Sigma;}}_{j=m-k+1}^{m-1}{(F_{i,j}^{\&prime;}-{\stackrel{\&OverBar;}{F}}_i^{\&prime;})}^2{\mathrm{\&Sigma;}}_{j=m-k+1}^{m-1}{(F_j^{\&prime;}-{\stackrel{\&OverBar;}{F}}^{\&prime;})}^2}}---\left(1\right)$

In formula (1), ${\stackrel{\&OverBar;}{F}}_i^{\&prime;}={\mathrm{\&Sigma;}}_{j=m-k+1}^{m-1}{F}_{i,j}^{\&prime;}/(k-1),$ ${\stackrel{\&OverBar;}{F}}^{\&prime;}={\mathrm{\&Sigma;}}_{j=m-k+1}^{m-1}{F}_j^{\&prime;}/(k-1);$

STEP2.2. to correlation coefficient r _i(i ∈ [1..n]) carries out following zero correction, obtains one group of weight and adjusts coefficient r ' _i, r ' _iShow more greatly forecast model f _iHigher with the actual result correlativity;

$r_i^{\&prime;}=\left|r_i\right|-{\mathrm{\&Sigma;}}_{i=1}^n\left|r_i\right|/n;---\left(2\right)$

STEP3. calculate forecast model f _i(i ∈ [1..n]) the m time result of calculation F ' _i,mWith actual result F ' _mDeviation Δ F _i:

STEP3.1. $\mathrm{\&Delta;}\stackrel{\&OverBar;}{F}=\frac{{\mathrm{\&Sigma;}}_{i=1}^n|F_{i,m}^{\&prime;}-F_m^{\&prime;}|}{n};---\left(3\right)$

STEP3.2. $\mathrm{\&Delta;}{F}_i=\mathrm{\&Delta;}\stackrel{\&OverBar;}{F}-|F_{i,m}^{\&prime;}-F_m^{\&prime;}|;---\left(4\right)$

Deviation Δ F _iLarger, show forecast model f _iThe m time forecasting accuracy high.

STEP4. calculate forecast model f _iThe weight adjusted value Δ ω of (i ∈ [1..n]) _i,m:

STEP4.1. to correlation coefficient r ' _i(i ∈ [1..n]) and deviation Δ F _iSpan normalization:

$r_i^{\&prime;}=\left|r_i\right|-{\mathrm{\&Sigma;}}_{i=1}^n\left|r_i\right|/n(r_i^{\&prime;}\&Element;[-\mathrm{1,1}\left]\right);---\left(5\right)$

$\mathrm{\&Delta;}{F}_i^{\&prime;}=\frac{\mathrm{\&Delta;}{F}_i+\min \left(\mathrm{\&Delta;}{F}_i\right)}{\max \left(\mathrm{\&Delta;}{F}_i\right)-\min \left(\mathrm{\&Delta;}{F}_i\right)}(\mathrm{\&Delta;}{F}_i^{\&prime;}\&Element;[-\mathrm{1,1}\left]\right);---\left(6\right)$

STEP4.2. calculate forecast model f _iThe weight adjusted value Δ ω of (i ∈ [1..n]) _i,m

$\mathrm{\&Delta;}{\mathrm{\&omega;}}_{i,m}={\mathrm{\&omega;}}_{i,m-1}\frac{(p_1{r}_i^{\&prime;}+{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="HDA00003472719200011.png"\; state="\{\{state\}\}">p</figure-callout>}}_2\mathrm{\&Delta;}{F}_i^{\&prime;})-{\mathrm{\&Sigma;}}_{i=1}^n(p_1{r}_i^{\&prime;}+{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="HDA00003472719200011.png"\; state="\{\{state\}\}">p</figure-callout>}}_2\mathrm{\&Delta;}{F}_i^{\&prime;})/n}{{\mathrm{\&Sigma;}}_{i=1}^n(p_1{r}_i^{\&prime;}+{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="HDA00003472719200011.png"\; state="\{\{state\}\}">p</figure-callout>}}_2\mathrm{\&Delta;}{F}_i^{\&prime;})}---\left(7\right)$

In formula (7), p ₁And p ₂Be respectively correlation coefficient r ' _i, deviation Δ F _iWeight, p wherein ₁+ p ₂=1 (p under default setting ₁=0.4p ₂=0.6).

STEP5., for each i ∈ [1..n], calculate ω _i,m;

ω′ _i,m=ω _i,m-1+Δω _i,mi∈[1..n] (8)

${\mathrm{\&omega;}}_{i,m}=\frac{{\mathrm{\&omega;}}_{i,m}^{\&prime;}}{{\mathrm{\&Sigma;}}_{i=1}^n{\mathrm{\&omega;}}_{i,m}^{\&prime;}}---\left(9\right)$

Weight adjusted value Δ ω _i,mMeet

Return to (ω _{1, m}, ω _{2, m}.., ω _i,m.., ω _n,m), algorithm withdraws from.

Claims (4)

1. the self-adapting regulation method of a linear combination forecasting model weight, is characterized in that, in accordance with the following steps:

(1) determine initial weight: establishing the numerical value of actual result first is F ' ₀, each forecast model f _iPredicted value be F ' _i; , for each i ∈ [1..n], calculate

Each forecast model f _iInitial weight be Meet the model that prediction effect is good first and have higher initial weight, and

(2) dynamically adjust weight:

A, establish current total m time and predict the outcome and actual result, comprise except the m time before the F ' that predicts the outcome of k-1 time _{I, m-k+1}, F ' _{I, m-k+2}.., R ' _{I, m-1}, and the actual result F ' of front k time _M-k+1, F ' _M-k+2.., F ' _m; Wherein, k is time window, defaultly gets 5; When m＜k, get k=m; Corresponding F ' _{I, m-k+1}, F ' _{I, m-k+2}.., F ' _{I, m-1}Weight be respectively w _{I, m-k+1}, w _{I, m-k+2}.., w _{I, m-1};

B, forecast model f _i, i ∈ [1..n], the result of calculation of front k-1 time and the related coefficient of front k actual result except the m time;

C, forecast model f _iThe deviation of the m time result of calculation and actual result,

D, according to the result of A, B, self-adaptation solves each forecast model f _iThe weights omega that (i ∈ [1..n) adjusts for the m time _i,m,, for high with the actual result correlativity and less with actual result deviation forecast model, increase the value of weight; For with the low forecast model of actual result correlativity and the forecast model larger with the actual result deviation, reduce the value of weight.

2. the self-adapting regulation method of linear combination forecasting model weight as claimed in claim 1, is characterized in that, in described step (2), the specific algorithm of B step is:

STEP1. calculate $F_m={\mathrm{\&Sigma;}}_{i=1}^n{\mathrm{\&omega;}}_{i,m-1}{F}_{i,m}^{\&prime;};$

Too frequently cause shake for fear of the weight adjustment, as F ' _m=F _m=0 or | F ' _m-F _m|/(| F ' _m|+| F _m|)≤5%, do not carry out the index weights adjustment, namely, for each i ∈ [1..n], make ω _i,m=ω _{I, m-1}, return to (ω _{1, m}, ω _{2, m}.., ω _i,m.., ω _n,m), algorithm withdraws from;

STEP2. according to forecast model f _iThe result of calculation of (i ∈ [1..n]) and the related coefficient of actual result, determine weight adjustment coefficient:

STEP2.1., for each i ∈ [1..n], adopt Pearson product-moment correlation coefficient (PPMCC) sequence of calculation F ' _{I, m-k+1}, F ' _{I, m-k+2}.., F ' _{I, m-1}With F ' _M-k+1, F ' _M-k+2.., F ' _m-1Related coefficient, obtain one group of correlation coefficient r _i, r _i∈ [1,1];

$r_i=\frac{{\mathrm{\&Sigma;}}_{j=m-k+1}^{m-1}(F_{i,j}^{\&prime;}-{\stackrel{\&OverBar;}{F}}_i^{\&prime;})(F_j^{\&prime;}-{\stackrel{\&OverBar;}{F}}^{\&prime;})}{\sqrt{{\mathrm{\&Sigma;}}_{j=m-k+1}^{m-1}{(F_{i,j}^{\&prime;}-{\stackrel{\&OverBar;}{F}}_i^{\&prime;})}^2{\mathrm{\&Sigma;}}_{j=m-k+1}^{m-1}{(F_j^{\&prime;}-{\stackrel{\&OverBar;}{F}}^{\&prime;})}^2}}---\left(1\right)$

In formula (1), ${\stackrel{\&OverBar;}{F}}_i^{\&prime;}={\mathrm{\&Sigma;}}_{j=m-k+1}^{m-1}{F}_{i,j}^{\&prime;}/(k-1),$ ${\stackrel{\&OverBar;}{F}}^{\&prime;}={\mathrm{\&Sigma;}}_{j=m-k+1}^{m-1}{F}_j^{\&prime;}/(k-1);$

STEP2.2. to correlation coefficient r _i(i ∈ [1..n]) carries out following zero correction, obtains one group of weight and adjusts coefficient r ' _i, r ' _iShow more greatly forecast model f _iHigher with the actual result correlativity;

$r_i^{\&prime;}=\left|r_i\right|-{\mathrm{\&Sigma;}}_{i=1}^n\left|r_i\right|/n;---\left(2\right)$

3. the self-adapting regulation method of linear combination forecasting model weight as claimed in claim 1, is characterized in that, in described step (2), the specific algorithm of C step is:

Calculate forecast model f _i(i ∈ [1..n]) the m time result of calculation F ' _i,mWith actual result F ' _m, deviation Δ F _i:

$\mathrm{\&Delta;}\stackrel{\&OverBar;}{F}=\frac{{\mathrm{\&Sigma;}}_{i=1}^n|F_{i,m}^{\&prime;}-F_m^{\&prime;}|}{n};---\left(3\right)$

$\mathrm{\&Delta;}{F}_i=\mathrm{\&Delta;}\stackrel{\&OverBar;}{F}-|F_{i,m}^{\&prime;}-F_m^{\&prime;}|;---\left(4\right)$

Deviation Δ F _iLarger, show forecast model f _iThe m time forecasting accuracy high.

4. the self-adapting regulation method of linear combination forecasting model weight as claimed in claim 1, is characterized in that, in described step (2), the specific algorithm of D step is:

STEP1. calculate forecast model f _iThe weight adjusted value Δ ω of (i ∈ [1..n]) _i,m:

STEP1.1. to correlation coefficient r ' _i(i ∈ [1..n]) and deviation Δ F _iSpan normalization:

$r_i^{\&prime;}=\left|r_i\right|-{\mathrm{\&Sigma;}}_{i=1}^n\left|r_i\right|/n(r_i^{\&prime;}\&Element;[-\mathrm{1,1}\left]\right);---\left(5\right)$

$\mathrm{\&Delta;}{F}_i^{\&prime;}=\frac{\mathrm{\&Delta;}{F}_i+\min \left(\mathrm{\&Delta;}{F}_i\right)}{\max \left(\mathrm{\&Delta;}{F}_i\right)-\min \left(\mathrm{\&Delta;}{F}_i\right)}(\mathrm{\&Delta;}{F}_i^{\&prime;}\&Element;[-\mathrm{1,1}\left]\right);---\left(6\right)$

STEP1.2. calculate forecast model f _iThe weight adjusted value Δ ω of (i ∈ [1..n]) _{I, m}

$\mathrm{\&Delta;}{\mathrm{\&omega;}}_{i,m}={\mathrm{\&omega;}}_{i,m-1}\frac{(p_1{r}_i^{\&prime;}+p_2\mathrm{\&Delta;}{F}_i^{\&prime;})-{\mathrm{\&Sigma;}}_{i=1}^n(p_1{r}_i^{\&prime;}+p_2\mathrm{\&Delta;}{F}_i^{\&prime;})/n}{{\mathrm{\&Sigma;}}_{i=1}^n(p_1{r}_i^{\&prime;}+p_2\mathrm{\&Delta;}{F}_i^{\&prime;})}---\left(7\right)$

In formula (7), p ₁And p ₂Be respectively correlation coefficient r ' _i, deviation Δ F _iWeight, p wherein ₁+ p ₂=1, p under default setting ₁=0.4p ₂=0.6;

STEP2., for each i ∈ [1..n], calculate ω _i,m;

ω′ _i,m=ω _i,m-1+Δω _i,mi∈[1..n] (8)

${\mathrm{\&omega;}}_{i,m}=\frac{{\mathrm{\&omega;}}_{i,m}^{\&prime;}}{{\mathrm{\&Sigma;}}_{i=1}^n{\mathrm{\&omega;}}_{i,m}^{\&prime;}}---\left(9\right)$

Weight adjusted value Δ ω _i,m, meet

Return to (ω _{1, m}, ω _{2, m}.., ω _i,m.., ω _n,m).

Images