No label defined (Q4336463)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 1008691
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | No label defined |
scientific article; zbMATH DE number 1008691 |
Statements
10 November 1997
0 references
ecological prediction
0 references
least squares
0 references
modeling
0 references
Grey Model
0 references
0.8378683924674988
0 references
0.8353865742683411
0 references
0.8271833062171936
0 references
0.8145291209220886
0 references
This paper deals with the improvement of the Grey Model \(GM(1,1)\) of \({\mathbf x}^{(1)}= [x^{(1)}_1,x^{(1)}_2,\dots, x^{(1)}_n]\) in the following whitened form \[ {d{\mathbf x}^{(1)}\over dt}= a{\mathbf x}^{(1)}+ u,\tag{1} \] where \(x^{(1)}_k= x^{(0)}_1+ x^{(0)}_2+\cdots+ x^{(0)}_k\), \(k= 1,2,\dots,n\). When \({\mathbf x}^{(0)}= [x^{(0)}_1,x^{(0)}_2,\dots, x^{(0)}_n]\) is given, using the least squares (LS) method leads to \[ \left[\begin{smallmatrix} a\\ u\end{smallmatrix}\right]= (B^\tau B)^{-1}B^\tau{\mathbf x},\quad B=\left[\begin{smallmatrix} b_2 & b_3 &\cdots & b_n\\ 1 & 1 &\cdots & 1\end{smallmatrix}\right]^\tau,\tag{2} \] where \({\mathbf x}= [x^{(0)}_2, x^{(0)}_3,\dots, x^{(0)}_n]^\tau\), \(b_k={1\over 2}[x^{(1)}_{k-1}+ x^{(1)}_k]\), \(k=2,3,\dots,n\). Based on \(\widehat x^{(1)}(k)\), the discrete solution to equation (1) satisfying \(\widehat x^{(1)}(1)= x^{(1)}_1\), the fitted values of all the components in \({\mathbf x}^{(0)}\) can be gained by \[ \widehat x^{(0)}(1)= x^{(1)}_1\text{ and }\widehat x^{(0)}(k)= \widehat x^{(1)}(k)-\widehat x^{(1)}(k- 1)= [x^{(0)}_1+\textstyle{{u\over a}}](1- e^{-a}),\;k\geq 2.\tag{3} \] For \(k>n\), \(\widehat x^{(0)}(k)\) is just the prediction, which has applications in the ecological area. The solution (3) is independent of \(x^{(0)}_1\), and an unknown ``suitable constant'' \(m\) should be added to each component in \({\mathbf x}^{(0)}\) so as to get a high modeling accuracy. An improved Grey Model \(GM^*(1,1)\) with higher fitting accuracy is further defined in the following whitened form: \[ {d{\mathbf x}^{(1)}\over dt}= a{\mathbf x}^{(1)}+ \beta t+\gamma.\tag{4} \] When \({\mathbf x}^{(0)}\) may be fitted by a series \(s^{(0)}_1,s^{(0)}_2,\dots, s^{(0)}_n\) satisfying \(s^{(0)}_k= Ee^{ak}+ F\), \(k=1,2,\dots,n\), (where \(a\) \(E\) and \(F\) are all constant), the LS method leads to \[ \left[\begin{matrix} \alpha\\ \beta\\ \gamma\end{matrix}\right]= (A^\tau A)^{-1}A^\tau{\mathbf x},\quad A=\left[\begin{matrix} b_2 &b_3 &\cdots &b_n\\ {3\over 2} &{5\over 2} &\cdots & {2n-1\over 2}\\ 1 & 1 &\cdots & 1\end{matrix}\right]^\tau.\tag{5} \] Then, the fitted values for \(x^{(0)}_k\)'s are: \(\widehat x^{(0)}(1)= x^{(1)}_1\) and \[ \widehat x^{(0)}(k)= \widehat x^{(1)}(k)- \widehat x^{(1)}(k- 1)= \Biggl[x^{(0)}_1+ {\beta+\gamma\over a}+ {\beta\over a^2}\Biggr]\Biggl(1-{1\over e^a}\Biggr) e^{a(k-1)}-{\beta\over a},\;k\geq 2\tag{6} \] and the ``suitable constant'' has been proved to be \(m=\beta/a\).
0 references