{"id":710,"date":"2020-12-20T10:48:35","date_gmt":"2020-12-20T01:48:35","guid":{"rendered":"http:\/\/cedartrees.co.kr\/?p=710"},"modified":"2021-04-03T19:09:27","modified_gmt":"2021-04-03T10:09:27","slug":"random-forest","status":"publish","type":"post","link":"http:\/\/blog.cedartrees.co.kr\/index.php\/2020\/12\/20\/random-forest\/","title":{"rendered":"\ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8(Random Forest)"},"content":{"rendered":"\n

\ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8\ub294 \uc758\uc0ac\uacb0\uc815\ud2b8\ub9ac(Decision Tree)\uc640 \ub2ee\uc740 \uc810\uc774 \ub9ce\uc740 \uc9c0\ub3c4\ud559\uc2b5 \uc608\uce21\ubaa8\ub378\uc785\ub2c8\ub2e4. \ub450 \ubaa8\ub378 \ubaa8\ub450 \uc5b4\ub5a4 \uc9c8\ubb38\uc5d0 \uc758\ud574\uc11c \ub370\uc774\ud130 \uc14b\uc744 \ubd84\ub9ac\ud558\ub294(\uac00\uc9c0\ub97c \ub9cc\ub4dc\ub294) \uae30\ubcf8\uc801\uc778 \ubc29\uc2dd\uc740 \ub2ee\uc558\uc9c0\ub9cc \uc758\uc0ac\uacb0\uc815\ub098\ubb34\uac00 \uc804\uccb4 \ub370\uc774\ud130\ub97c \ud1b5\ud574\uc11c \uc131\ub2a5\uc774 \uc88b\uc740 \ud558\ub098\uc758 \ub098\ubb34\ub97c \ub9cc\ub4dc\ub294\ub370 \ubaa9\uc801\uc774 \uc788\ub2e4\uba74 \ub7a8\ub364\ud3ec\ub808\uc2a4\ud2b8\ub294 \ud558\ub098\uc758 \ub098\ubb34\ub97c \ub9cc\ub4e4\uae30 \ubcf4\ub2e4\ub294 \ub370\uc774\ud130\uc14b\uc744 \ub79c\ub364\ud558\uac8c \uc0d8\ud50c\ub9c1\ud574\uc11c \uc5ec\ub7ec\uac1c\uc758 \uc608\uce21 \ubaa8\ub378\uc744 \ub9cc\ub4e4\uace0 \uadf8 \uc774\ub7ec\ud55c \ubaa8\ub378\ub4e4\uc744 \uc885\ud569\ud574\uc11c \ud558\ub098\uc758 \uc608\uce21 \uacb0\uacfc\ub97c \ub9ac\ud134\ud558\ub294 \ubc29\ubc95\uc73c\ub85c \ucd5c\uc885 \uc608\uce21\uc744 \uc218\ud589\ud569\ub2c8\ub2e4.<\/p>\n\n\n\n

\uc774\ub7ec\ud55c \uacfc\uc815\uc744 Bagging(= Bootstrap + Aggregation)\uc774\ub77c\uace0 \ud569\ub2c8\ub2e4. \uc989, \uc8fc\uc5b4\uc9c4 \ud558\ub098\uc758 \ud070 \ub370\uc774\ud130\ub97c \uc5ec\ub7ec \uac1c\uc758 \ubd80\ud2b8\uc2a4\ud2b8\ub7a9 \uc790\ub8cc(\uc911\ubcf5\ud5c8\uc6a9)\ub97c \uc0dd\uc131\ud558\uace0 \uac01 \ubd80\ud2b8\uc2a4\ud2b8\ub7a9 \uc790\ub8cc\ub97c \ubaa8\ub378\ub9c1\ud55c \uacb0\uacfc\ub97c \ud1b5\ud569\ud558\uc5ec \ucd5c\uc885\uc758 \uc608\uce21 \ubaa8\ub378\uc744 \uc0b0\ucd9c\ud558\ub294 \ubc29\ubc95\uc785\ub2c8\ub2e4. \uc774\ub7f0 \uc608\uce21 \ubaa8\ub378\uc740 \ub370\uc774\ud130\uac00 \ubcc0\ub3d9\uc131\uc774 \ud070 \uacbd\uc6b0 \uc6d0\uc790\ub8cc(Raw Data)\ub85c\ubd80\ud130 \uc5ec\ub7ec\ubc88\uc758 \uc0d8\ud50c\ub9c1\uc755 \ud1b5\ud574\uc11c \uc608\uce21 \ubaa8\ub378\uc758 \uc815\ud655\ub3c4\ub97c \ub192\uc774\ub294 \uae30\ubc95\uc785\ub2c8\ub2e4. \uc774\ub7f0 \uacfc\uc815\uc744 \ud1b5\ud574\uc11c n\uac1c\uc758 \uc608\uce21 \ubaa8\ub378\uc774 \ub9cc\ub4e4\uc5b4\uc9d1\ub2c8\ub2e4. \uadf8\ub7ec\ub098 \uacb0\uad6d \ud544\uc694\ud55c \uac83\uc740 \ud558\ub098\uc758 \ub370\uc774\ud130\uc774\uae30 \ub54c\ubb38\uc5d0 \uac01\uac01\uc758 \ubaa8\ub378\uc758 \uacb0\uacfc \uac12\uc5d0 \ub300\ud574\uc11c \ud68c\uadc0\ubd84\uc11d(\ud3c9\uade0\uacc4\uc0b0)\uc744 \ud558\ub358\uac00 \uacfc\ubc18 \ud22c\ud45c(\ubd84\ub958 \ubaa8\ub378)\ub97c \ud1b5\ud574\uc11c \ucd5c\uc885 \uacb0\uacfc \uac12\uc744 \ub9cc\ub4e4\uac8c \ub429\ub2c8\ub2e4.
\uc774\ub7ec\ud55c \ubc29\ubc95\uc758 \uc54c\uace0\ub9ac\uc998 \uc911\uc5d0 \uac00\uc7a5 \ub300\ud45c\uc801\uc778 \ubaa8\ub378\uc774 \ubc14\ub85c \ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8(Random Forest)\uc785\ub2c8\ub2e4.<\/p>\n\n\n\n

\uc774\ub7ec\ud55c \ubcf5\uc7a1\ud55c \uc54c\uace0\ub9ac\uc998\uc744 sklearn\uc758 ensemble \ud328\ud0a4\uc9c0\uc758 RandomForestClassifier\uc5d0\uc11c \ud6cc\ub96d\ud558\uac8c \uad6c\ud604\ud558\uace0 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n

import numpy as np\nimport pandas as pd\n\nfrom sklearn.datasets import make_blobs, make_classification\nfrom sklearn.model_selection import train_test_split\nimport matplotlib.pyplot as plt<\/pre>\n\n\n\n
\ud14c\uc2a4\ud2b8 \ub370\uc774\ud130 \uc0dd\uc131\uc744 \uc704\ud574\uc11c sklearn\uc758 make_classification\ub97c \uc0ac\uc6a9\ud558\uaca0\uc2b5\ub2c8\ub2e4. \ub370\uc774\ud130\uc758 \uc218\ub294 \ucd1d 300\uac1c, feature\ub294 5\uac1c\uc774\uba70\uc774\uba70 \uac01\uac01\uc758  feature\ub294 \uc885\uc18d\ubcc0\uc218\uc640 \uc0c1\uad00\uad00\uacc4\uac00 \uc874\uc7ac\ud569\ub2c8\ub2e4. \ud574\ub2f9 \ub370\uc774\ud130 \uc14b\uc758 \ud074\ub7ec\uc2a4\ud130\ub294 1\uc774\uace0 \ub370\uc774\ud130\uc758 \ud074\ub798\uc2a4\ub294 3\uc785\ub2c8\ub2e4. \uc989, \ucd5c\uc885 \uc608\uce21\uc740 0,1,2 \uc774 \uc14b \uc911\uc5d0 \ud558\ub098\uc758 \uac12\uc744 \uac00\uc9c4\ub2e4\ub294 \ub73b\uc785\ub2c8\ub2e4.<\/p>\n\n\n\n
\uc704\uc5d0 \ubcf4\uba74 \ub4f1\ubd84\uc0b0\uc131 \ub370\uc774\ud130\ub97c \ub9cc\ub4e4\uc5b4\uc8fc\ub294 make_blobs()\ub97c \uc0ac\uc6a9\ud55c \ubd80\ubd84\uc774 \uc788\ub294\ub370 \uc774\ub807\uac8c \ud558\uba74 \ub370\uc774\ud130\uac00 \ubd84\ub958\uc5d0 \ub108\ubb34 \ucd5c\uc801\ud654 \ub418\uc5b4 \uc788\uae30 \ub54c\ubb38\uc5d0 make_classification()\uc744 \uc0ac\uc6a9\ud569\ub2c8\ub2e4.<\/p>\n\n\n\n
#x, y = make_blobs(n_samples=300, n_features=5, centers=3) \nx, y = make_classification(n_samples=300, n_features=5, n_informative=5, n_redundant=0, n_clusters_per_class=1, n_classes=3)<\/pre>\n\n\n\n
\ub370\uc774\ud130 \uc14b\uc744 \ub9cc\ub4e0 \ub2e4\uc74c \ubaa8\ub378\uc758 \uc815\ud655\ub3c4\ub97c \uac80\uc99d\ud558\uae30 \uc704\ud574\uc11c train, test \ub370\uc774\ud130 \uc14b\uc73c\ub85c \ubd84\ub9ac\ud569\ub2c8\ub2e4. \ubd84\ub9ac\ub294 8:2 \uc815\ub3c4\ub85c \ud558\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n
x_train, x_test, y_train, y_test = train_test_split(x, y, test_size = 0.2, random_state=0) # 8:2\nprint(x_train.shape, x_test.shape)\n# (240, 5) (60, 5)<\/pre>\n\n\n\n
sklearn\uc5d0 \ubcf4\uba74 RandomForestClassifier\ub97c \ub9cc\ub4dc\ub294\ub370 \uba87\uac00\uc9c0 \ud30c\ub77c\uba54\ud130\ub4e4\uc774 \uc788\uc2b5\ub2c8\ub2e4. \uadf8\uc911\uc5d0 \uc608\uc81c\uc5d0\uc11c\ub294 n_estimators(The number of trees in the forest.)\ub97c \uc0ac\uc6a9\ud569\ub2c8\ub2e4. \ud574\ub2f9 \ud30c\ub77c\uba54\ud130\ub294 \ud558\ub098\uc758 \ub098\ubb34\ub97c \ub9cc\ub4dc\ub294 \uc758\uc0ac\uacb0\uc815\ub098\ubb34(Decision-Tree)\uc640\ub294 \ub2e4\ub974\uac8c \uc5ec\ub7ec\uac1c\uc758 \ubaa8\ub378\uc744 \ub9cc\ub4dc\ub294 RandomForest\uc758 \ud2b9\uc9d5\uc785\ub2c8\ub2e4. \ub098\ubb34\ub97c \ub9ce\uc774 \ub9cc\ub4e4\uba74 \uc608\uce21\uc758 \uc815\ud655\ub3c4\ub294 \ub192\uc544 \uc9c8 \uc218 \uc788\uc9c0\ub9cc \uadf8\ub9cc\ud07c \ub9ce\uc740 \uc790\uc6d0\uc744 \ud544\uc694\ub85c \ud558\uace0 \ub610 \uc77c\uc815 \uc218\uc900 \uc774\uc0c1\uc73c\ub85c\ub294 \ub192\uc544\uc9c0\uc9c0 \uc54a\uae30 \ub54c\ubb38\uc5d0 \uac00\ub2a5\ud558\uba74 \ud14c\uc2a4\ud2b8\ub97c \ud558\uba74\uc11c \uc218\ub97c \ub298\ub824\uac00\ub294 \ubc29\uc2dd\uc73c\ub85c \uc218\ud589\ud558\ub294 \uac83\uc744 \ucd94\ucc9c\ud569\ub2c8\ub2e4.<\/p>\n\n\n\n
\ud574\ub2f9 \ubaa8\ub378\uc740 \ub370\uc774\ud130\uc14b\uc774 \ube44\uad50\uc801 \uad6c\ubd84\uc774 \uc798\ub418\uc5b4 \uc788\ub294 \ub370\uc774\ud130\uc14b\uc774\uae30 \ub54c\ubb38\uc5d0 10\uc73c\ub85c \uc14b\ud305\ud569\ub2c8\ub2e4.<\/p>\n\n\n\n
from sklearn.ensemble import RandomForestClassifier\nrandomfc = RandomForestClassifier(n_estimators=10).fit(x_train, y_train)\nrandomfc<\/pre>\n\n\n\n
RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n                       max_depth=None, max_features='auto', max_leaf_nodes=None,\n                       min_impurity_decrease=0.0, min_impurity_split=None,\n                       min_samples_leaf=1, min_samples_split=2,\n                       min_weight_fraction_leaf=0.0, n_estimators=10,\n                       n_jobs=None, oob_score=False, random_state=None,\n                       verbose=0, warm_start=False)<\/pre>\n\n\n\n
\ubaa8\ub378\uc758 \uc815\ud655\ub3c4\ub294 85%\ub85c \uce21\uc815\ub418\uc5c8\uc2b5\ub2c8\ub2e4. \ub370\uc774\ud130 \uc14b\uc740 \ub79c\ub364\uc73c\ub85c \ub9cc\ub4e4\uc5b4\uc9c0\uae30 \ub54c\ubb38\uc5d0 \ub9cc\ub4e4 \ub54c\ub9c8\ub2e4 \uc815\ud655\ub3c4\uac00 \ub2e4\ub974\uac8c \uce21\uc815\ub429\ub2c8\ub2e4.<\/p>\n\n\n\n
predict = randomfc.predict(x_test)\nprint(randomfc.score(x_test, y_test))\n# 0.85<\/pre>\n\n\n\n
\ub354 \uc790\uc138\ud55c \uacb0\uacfc\ub97c \ubcf4\uae30 \uc704\ud574\uc11c confusion_matrix\ub97c \uccb4\ud06c\ud574\ubd05\ub2c8\ub2e4. 2~4\uac1c\uc758 \uc624\ucc28\uac00 \ubc1c\uacac\ub418\uc9c0\ub9cc\uacb0\uacfc \uad00\ub300\ud55c \uc800\uc5d0\uac8c\ub294 \ub300\ubd80\ubd84 \uc815\ud655\ud558\uac8c \uc608\uce21\uc744 \ud55c\uac83\uc73c\ub85c \ubcf4\uc785\ub2c8\ub2e4. (\ub2e4\ub974\uac8c \ubcf4\uc77c \uc218\ub3c4 \uc788\uc2b5\ub2c8\ub2e4.)<\/p>\n\n\n\n
from sklearn.metrics import confusion_matrix\nconfusion_matrix(y_test, predict)\n# output #\narray([[14,  1,  3],\n       [ 1, 18,  1],\n       [ 3,  0, 19]])<\/pre>\n\n\n\n
classification_report\ub97c \ud1b5\ud574\uc11c \uacb0\uacfc\ub97c \ub354 \uc790\uc138\ud788 \uc0b4\ud3b4\ubcf4\uaca0\uc2b5\ub2c8\ub2e4. \uac01\uac01\uc758 \uc6a9\uc5b4\ub4e4\uc740 \uc544\ub798\uc640 \uac19\uc740 \uc758\ubbf8\uac00 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n
\uc815\ud655\ub3c4(accuracy) : \uc608\uce21\ud55c \uac12\uc758 \uba87\uac1c\ub97c \ub9de\ucdc4\ub294\uac00?
\uc815\ubc00\ub3c4(precision) : \uc608\uce21\ud55c \uac83\uc911\uc5d0 \uc815\ub2f5\uc758 \ube44\uc728\uc740?
\uc7ac\ud604\uc728(recall) : \uc815\ub2f5\uc778 \uac83\uc744 \ubaa8\ub378\uc774 \uc5b4\ub5bb\uac8c \uc608\uce21\ud588\ub294\uac00? 
F1 Score : \uc815\ubc00\ub3c4\uc640 \uc7ac\ud604\uc728\uc758 \uac00\uc911\uc870\ud654\ud3c9\uade0(weight harmonic average)\uc744 F\uc810\uc218(F-score)\ub77c\uace0 \uc815\uc758\ud569\ub2c8\ub2e4. \uc989, F1 Score \uac12\uc774 \ub192\uc73c\uba74 \uc131\ub2a5\uc774 \ub192\ub2e4\uace0 \ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n
from sklearn.metrics import classification_report\nreport = classification_report(y_test, predict)\n# output #\n              precision    recall  f1-score   support\n\n           0       0.78      0.78      0.78        18\n           1       0.95      0.90      0.92        20\n           2       0.83      0.86      0.84        22\n\n    accuracy                           0.85        60\n   macro avg       0.85      0.85      0.85        60\nweighted avg       0.85      0.85      0.85        60<\/pre>\n\n\n\n
\uc774\uc81c \ub79c\ub364\ud558\uac8c \uc0dd\uc131\ub41c \uc784\uc758\uc758 \ub370\uc774\ud130\uac00 \uc544\ub2cc load_digits \ub370\uc774\ud130\uc14b\uc744 \ud65c\uc6a9\ud574\ubcf4\ub3c4\ub85d \ud558\uaca0\uc2b5\ub2c8\ub2e4. \ud574\ub2f9 \ub370\uc774\ud130\uc14b\uc5d0 \ub300\ud55c \uc124\uba85\uc740 \uc0dd\ub7b5\ud558\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n
from sklearn.datasets import load_digits\ndigits = load_digits()\n\nx_train, x_test, y_train, y_test = train_test_split(digits.data, digits.target, test_size = 0.2, random_state=0) # 8:2\nprint(x_train.shape, x_test.shape)\n# (1437, 64) (360, 64)<\/pre>\n\n\n\n
randomfc = RandomForestClassifier(n_estimators=10).fit(x_train, y_train)\nprint(randomfc.score(x_test, y_test) )\n# 0.9472222222222222<\/pre>\n\n\n\n
predict = randomfc.predict(x_test)\nconfusion_matrix(y_test, predict)\n\narray([[27,  0,  0,  0,  0,  0,  0,  0,  0,  0],\n       [ 0, 34,  0,  0,  0,  1,  0,  0,  0,  0],\n       [ 1,  0, 33,  2,  0,  0,  0,  0,  0,  0],\n       [ 0,  0,  0, 29,  0,  0,  0,  0,  0,  0],\n       [ 0,  0,  0,  0, 28,  0,  0,  2,  0,  0],\n       [ 0,  0,  0,  0,  0, 39,  0,  0,  1,  0],\n       [ 0,  1,  0,  0,  0,  0, 42,  0,  1,  0],\n       [ 0,  0,  0,  0,  0,  0,  0, 38,  1,  0],\n       [ 0,  2,  0,  2,  0,  0,  0,  0, 35,  0],\n       [ 0,  3,  0,  1,  0,  0,  0,  1,  0, 36]])<\/pre>\n\n\n\n
report = classification_report(y_test, predict)\nprint(report)\n\n              precision    recall  f1-score   support\n\n           0       0.96      1.00      0.98        27\n           1       0.85      0.97      0.91        35\n           2       1.00      0.92      0.96        36\n           3       0.85      1.00      0.92        29\n           4       1.00      0.93      0.97        30\n           5       0.97      0.97      0.97        40\n           6       1.00      0.95      0.98        44\n           7       0.93      0.97      0.95        39\n           8       0.92      0.90      0.91        39\n           9       1.00      0.88      0.94        41\n\n    accuracy                           0.95       360\n   macro avg       0.95      0.95      0.95       360\nweighted avg       0.95      0.95      0.95       360<\/pre>\n\n\n\n
<\/p>\n","protected":false},"excerpt":{"rendered":"
\ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8\ub294 \uc758\uc0ac\uacb0\uc815\ud2b8\ub9ac(Decision Tree)\uc640 \ub2ee\uc740 \uc810\uc774 \ub9ce\uc740 \uc9c0\ub3c4\ud559\uc2b5 \uc608\uce21\ubaa8\ub378\uc785\ub2c8\ub2e4. \ub450 \ubaa8\ub378 \ubaa8\ub450 \uc5b4\ub5a4 \uc9c8\ubb38\uc5d0 \uc758\ud574\uc11c \ub370\uc774\ud130 \uc14b\uc744 \ubd84\ub9ac\ud558\ub294(\uac00\uc9c0\ub97c \ub9cc\ub4dc\ub294) \uae30\ubcf8\uc801\uc778 \ubc29\uc2dd\uc740 \ub2ee\uc558\uc9c0\ub9cc \uc758\uc0ac\uacb0\uc815\ub098\ubb34\uac00 \uc804\uccb4 \ub370\uc774\ud130\ub97c \ud1b5\ud574\uc11c \uc131\ub2a5\uc774 \uc88b\uc740 \ud558\ub098\uc758 \ub098\ubb34\ub97c \ub9cc\ub4dc\ub294\ub370 \ubaa9\uc801\uc774 \uc788\ub2e4\uba74 \ub7a8\ub364\ud3ec\ub808\uc2a4\ud2b8\ub294 \ud558\ub098\uc758 \ub098\ubb34\ub97c \ub9cc\ub4e4\uae30 \ubcf4\ub2e4\ub294 \ub370\uc774\ud130\uc14b\uc744 \ub79c\ub364\ud558\uac8c \uc0d8\ud50c\ub9c1\ud574\uc11c \uc5ec\ub7ec\uac1c\uc758 \uc608\uce21 \ubaa8\ub378\uc744 \ub9cc\ub4e4\uace0 \uadf8 \uc774\ub7ec\ud55c \ubaa8\ub378\ub4e4\uc744 \uc885\ud569\ud574\uc11c \ud558\ub098\uc758 \uc608\uce21 \uacb0\uacfc\ub97c \ub9ac\ud134\ud558\ub294 \ubc29\ubc95\uc73c\ub85c \ucd5c\uc885 \uc608\uce21\uc744 … <\/p>\n
\ub354 \ubcf4\uae30 “\ub79c\ub364\ud3ec\ub808\uc2a4\ud2b8(Random Forest)”<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[17,70],"tags":[54,101,48,102,45,55],"_links":{"self":[{"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/posts\/710"}],"collection":[{"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/comments?post=710"}],"version-history":[{"count":6,"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/posts\/710\/revisions"}],"predecessor-version":[{"id":718,"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/posts\/710\/revisions\/718"}],"wp:attachment":[{"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/media?parent=710"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/categories?post=710"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/blog.cedartrees.co.kr\/index.php\/wp-json\/wp\/v2\/tags?post=710"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}