Figure 5
From: Machine learning-aided analysis for complex local structure of liquid crystal polymers
Correct answer rate (i.e., accuracy) for combination of features selected by SFS. The accuracy increases with respect to N and the number of features. For N = 4, 5, and 6, an increase in the number of features up to four effectively improves the accuracy. The addition of the fifth feature does not improve the results. For N = 8, the accuracy is saturated with two or more features. For N = 12, the accuracy is almost ideal regardless of the number of features. The condition N = 8 with two features and the condition N = 6 with four features is close to the accuracy of the condition N = 12.
