Rivercan/feature_selection

feature selection by using random forest.

Feature_selection

use random forest to find important feature.

1. feature importance ranking:

• By Feature importance ranking top10

• 1. TB_9a
• 2. TB_a9
• 3. Img0.1
• 4. ent_p_5
• 5. TB_b1
• 6. TB_71
• 7. ent_p_8
• 8. TB_ce
• 9. GetStringTypeA
• 10. ExitProcess

2. Useless feature ranking:

• Featurn ranking list Top10 useless feature:

• GetLastActivePopup
• ImageList_Add
• GlobalDeleteAtom
• IsBadReadPtr
• SelectPalette
• GetMenuState
• ExitThread
• AdjustWindowRectEx
• GetEnvironmentVariableA
• SHGetFileInfoA

3. decision method

• 使用 sklearn 裡面的 RandonforestClassifier，取前10為 important features，後10為useless features。
• useless features 只代表相對不佳，不代表這些 useless features 對分類完全沒用，important features 亦同。

4. packages

• sklearn package 中的 RandomForestClassifier 做 feature selection
• pandas，numpy 處理資料
• matplotlib.pyplot 畫圖

5. suggestion

• 無