Appropriate data format for fitcsvm() function

2 次查看（过去 30 天）

Saugata Bose 2019-8-1

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/474482-appropriate-data-format-for-fitcsvm-function

编辑： Saugata Bose 2019-8-2

Im a trying to design a svm classifier. Doing this, I stuck at a point where I am getting an error during applying fitcsm(). I understand that, the error arises due to the mismatch of the appropriate data format. Would you please help me to sort out the issue?

Input:

I have 2000 train data of format '1 x 20 x 300 dimension single' which is X where as the category data for train is of 2000 x 1 dimension which is Y.

Data Format:

The data format of X{2000 X 1} is following

Each cell contains data of dimension 1 x 20 x 300 such as the following:

where as Y contains categorical data of 2000 x 1.

After applying fitcsvm(X, Y) the following error has arisen:

Invalid data type. X must be a double or single matrix.

Would you please suggest me the appropriate way to make fitcsvm() working?

thanks,

2 个评论
显示无隐藏无

the cyclist 2019-8-1

If you have 2000 Y data (size = [2000,1] as you say), then I would expect your X data to have size [2000,N], where N is the number of features in your feature set. I don't understand how a [1,20,300] array -- which has 6000 elements -- translates into the size you need.

Saugata Bose 2019-8-2

编辑：Saugata Bose 2019-8-2

@the cyclist:

Sorry for the delayed response. Actually, after preprocessing X it turns into a coplex dimension(I believe). Each X is of dimension '1 x 20 x 300 single'. I found that, fitcsvm expects X either be a table format or a single/double format. Here, I got stuck that how could i change the dimension in either table format or in single/doble format. After getting your reply, I have tried to elaborate my query by adding few images in the updated question as well.

请先登录，再进行评论。

请先登录，再回答此问题。