K-Means with their feature selection in RapidMiner

Saksit Jantaraplin

·

Follow

4 min read

·

Aug 31, 2021

--

Listen

Share

Related Research

Design

Without Feature Selection

• accuracy: 65.52%

• split data: 0.01, accuracy: 66.22%

Unsupervised Feature Selection with k-means

• balance of simplicity: 1.0 = accuracy: 63.23%

• balance of simplicity: 0.5 = accuracy: 63.23%

• split data: 0.01, balance of simplicity: 1.0 = accuracy: 57.72%

• split data: 0.01, balance of simplicity: 0.5 = accuracy: 66.22%

Try to use different feature selection model

K-Means (H2O)

• notsplit 0.5

• notsplit 1.0

• split 0.5

• split 1.0

X-Means

• notsplit 0.5

• notsplit 1.0

• split 0.5

• split 1.0

K-NN

• split max 20

• split max 30

Naive Bayes

• split max 20

• split max 30

Decission Tree

• split max 20

• split max 30