Top Page | Upper Page | Contents | About This Site | JAPANESE

Difference of Output by Methods

In Difference of Good Distribution by Methods, "favorite distributions" is the point of view.

In this page, "output" is the point of view. This is the important point when we use these methods for prediction.

Making This Page

I used MS-Excel to make sample data and graphs. And I used RapidMiner to use methods. Sample file is an example.

In this page, I do not use MT method because RapidMiner does not cover this method.

RapidMiner Part

The right figure is the part of RapidMiner.

• Read CSV : To read sample data. "Y" is changed into binominal and label.
• Read CSV(2) : To read the data for prediction.
  CSV file for prediction
• k-NN : To make model. This example is for k-NN. I used default value for each parameter for most part. Some values are changed.
• Apply Model : To predict

Graphs in This Page

Red points are "-1" in the sample data. And blue points are "1".

Light red parts are the area predicted as "-1". Light blue parts are as "1".

Methods

In RapidMiner, Kernel Function can be used for Logistic Regression Analysis. So the methods are not same explained in the page, Difference of Good Distribution by Methods.

• DA : Stand for "Discriminant Analysis"
• LR : Stand for "Logistic Regression Analysis"
  "kernel type" is "radial".
• SVM : Stand for "Support Vector Machine"
  "kernel type" is "radial".
• DT : Stand for "Decision Tree"
• k-NN : Stand for "k Nearest Neighbor"
  k = 3
• NN : Stand for "Neural Network"
• NB : Stand for "Naive Bayes"

Method and Graphs

The best methods are changed when purpose or data are changed.

Case 1

Link for input file

Case 2

Link for input file

Case 3

Link for input file

NEXT Analyzing variable companioning