libsvm的准确度为0

Question

我想问一下你对我的问题的看法。我有两类问题。这是我的程序：

1. 对数据集进行Z分数归一化
2. 降维由PCA
完成
3. 使用交叉验证
4. 我试图根据中的例子使用libsvm precomputed kernels with libsvm

我的准确率为0。那是什么原因？据我所知，两类问题的准确度不能低于50％。我有32x2967矩阵，其中包含我的数据集。前16个属于第一类，其余属于第二类。

这是令我担心的代码

-- Find "CurrentName" (without quotes) and replace with the current database name
-- Find "NewDbName" (without quotes) and replace with the new database name


USE [CurrentName];

-- Change physical file names:
ALTER DATABASE [CurrentName] MODIFY FILE (NAME = 'CurrentName', FILENAME = '<Full-Path-Required>\NewDbName.mdf');
ALTER DATABASE [CurrentName] MODIFY FILE (NAME = 'CurrentName_log', FILENAME = '<Full-Path-Required>\NewDbName_log.ldf');

-- Change logical names:
ALTER DATABASE [CurrentName] MODIFY FILE (NAME = 'CurrentName', NEWNAME = 'NewDbName');
ALTER DATABASE [CurrentName] MODIFY FILE (NAME = 'CurrentName_log', NEWNAME = 'NewDbName_log');

-- Take offline and back online
USE [master]
GO
ALTER DATABASE [CurrentName] SET OFFLINE WITH ROLLBACK IMMEDIATE;
-- Then navigate to <Full-Path-Required> and rename the files
ALTER DATABASE [CurrentName] SET ONLINE;