You are on page 1of 7

10 most significant variables

Linear Regression Model


Q= 108 -0.0050784 x1 + 0.69897 x2 -1.2359 x3 + 0.26047 x4 + 0.032026 x5 -0.040235 x6 -0.90058 x7 +
0.97637 x8 -0.093918 x9 -0.19729 x10

120
Qactual Qmodel
100

80

60

40

20

Root Mean Squared Error: 4.36

R-squared: 0.244, Adjusted R-Squared: 0.243

Non-linear Regression
System Identification
Fit to estimation data: 77.2% (prediction focus)

FPE: 1.31, MSE: 1.307

Neural Network
3 MOST SIGNIFICANT DATA
LINEAR MODEL
Multiple linear regression
Q=122.22-0.65621x1+0.15056x2-0.008022x3

Comparison of Qactual and Qmodel


120

100

80

60

40

20

0
45 1260 2475 3690 4905 6120 7335 8550 9765 109801219513410146251584017055182701948520700

Qactual Qmodel

Neural Network

Linear

NARX
Neural Network

Linear Performance

NARX Performance
Neural Network

Linear Regressors

NonLinear Regressors
NARX MODEL COMPARISON USING SYSTEM IDENTIFICATION

RED=wavelet based narx

BLUE=sigmoid based narx

Green= narx

Regressor

You might also like