Why Standardization = false is slow & has low accuracy?

2 次查看（过去 30 天）
Zulfiqar Khan
Zulfiqar Khan 2018-12-3
Hi,
I am working on SVM kernel using Matlab. I have got following examples:
Quadratic SVM accuracy is 7.3%; Kernel function = quadratic, Kernel Scale Mode = Auto; Multiclass method one versus all, Box constrained level = 4 Standardized data=false
Quadratic SVM accuracy 9.1% and confusion matrix mode auto multiclass method one versus all Box constraint 10 standarddize data false PCA disable
Quadratic SVM accuracy is 9.4%; Kernel function = quadratic, Kernel Scale Mode = Auto; Multiclass method one versus one, Box constrained level = 4 Standardized data=false 2960.2s
Quadratic SVM accuracy is 19.6; Kernel function = quadratic, Kernel Scale Mode = Manual; Multiclass method one versus one, Box constrained level = 4 Standardized data=false ,2682s
In all the above examples, it generated a low accuracy and kernel took lot of time. I found that we don't do normalization and encoding in Standardization= false. So what is the reason for this bad performance.
Good accuracy is 62.8 and better with Standardization= true for quadratic and Guassian kernels. Please guide me.
Zulfi.

请先登录，再回答此问题。

回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Get Started with Statistics and Machine Learning Toolbox 的更多信息

标签

产品

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by