3
2

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?

More than 5 years have passed since last update.

Spark mllib 教師あり学習アルゴリズム精度測定 - linear classifier編

Last updated at Posted at 2015-09-22

概要

spark mllibの

  • SVM
  • LogisticRegression
  • NaiveBayes

の精度測定を、

  • news20
  • rcv1

に対して実施。その結果を記載する。

各アルゴリズムのパラメータ設定は

Svm

  • L2正則化のみ
  • numIteration = 1000
  • SGD, miniBatchFraction = 1.0
  • regParam=c= 1.0* 10^-3 〜 1000まで測定

LogisticRegression (LR)

  • L2正則化のみ
  • numIteration = 1000
  • SGD, miniBatchFraction = 1.0
  • regParam=c= 1.0* 10^-3 〜 10まで測定

NaiveBayes (NB)

  • lambda=1.0*10^-3 〜 1000

測定結果

news20 NB

lambda 正例F1値 正解率
0.0001 0.975 0.975
0.001 0.974 0.974
0.01 0.973 0.972
0.1 0.967 0.967
1.0 0.931 0.933
10.0 0.838 0.855

rcv1 NB

lambda 正例F1値 正解率
0.0001 0.925 0.924
0.001 0.932 0.932
0.01 0.937 0.935
0.1 0.943 0.941
1.0 0.941 0.939
10.0 0.932 0.931

rcv1 LR

regParam 正例F1値 正解率
0.0001 0.924 0.922
0.001 0.924 0.922
0.01 0.924 0.922
0.1 0.924 0.922
1.0 0.924 0.922
10.0 0.924 0.922
100.0 NaN 0.481
1000.0 NaN 0.481

rcv1 svm

regParam 正例F1値 正解率
0.0001 0.924 0.922
0.001 0.924 0.922
0.01 0.924 0.922
0.1 0.924 0.922
1.0 0.924 0.922
10.0 0.924 0.922
100.0 NaN 0.480
1000.0 NaN 0.480
3
2
0

Register as a new user and use Qiita more conveniently

  1. You get articles that match your needs
  2. You can efficiently read back useful information
  3. You can use dark theme
What you can do with signing up
3
2

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?