Practical Applications of Data Mining

4.7 linear regreSSion 169

One useful measure of the adequacy of the fitted model is the coeffi-

cient of determination, R

2

= SSR/TSS = 1 – SSE/TSS, which lies between

0 and 1. It expresses the regression sum of squares as a percentage of the

TSS. A high R

2

value indicates that the regression model with n independent

attributes explains a large portion of the variability among the observations y

1

,

y

2

, …, y

m

. Hence, when the value of R

2

is close to 1, the predicted dependent

variable is closer to the observed dependent variable and the fitting is better.

Since this R

2

value by itself does not tell us which of the n variables are the

most impor ...