what is the cause of Accuracy difference?

Question

我有一个2001属性和63个实例的数据集，当使用Weka进行分类时，我们可以看到J48的准确性要比Naive Bayes高。但是在进行PCA之后，我们可以看到J48的准确性与第一个获得的准确性相比有所下降82％，PCA之后我获得了72.5％。这是什么原因？

Answer 1

Decision trees select features.

PCA mixes features into components.

This makes feature selection work much less - every input feature influences (almost) every derived feature.

Don't assume PCA is a magic tool that solves all your problems. It can work - but it can also make things worse.