集成学习方法之随机森林
集成学习方法之随机森林
什么是集成学习方法
集成学习通过建立几个模型组合的来解决单一预测问题。它的工作原理是生成多个分类器/模型,各自独立地学习和作出预测。这些预测最后结合成组合预测,因此优于任何一个单分类的做出预测。
什么是随机森林
在机器学习中,随机森林是一个包含多个决策树的分类器,并且其输出的类别是由个别树输出的类别的众数而定。
例如, 如果你训练了5个树, 其中有4个树的结果是True, 1个数的结果是False, 那么最终投票结果就是True
随机森林原理过程
学习算法根据下列算法而建造每棵树:
- 用N来表示训练用例(样本)的个数,M表示特征数目。
- 1、一次随机选出一个样本,重复N次, (有可能出现重复的样本)
- 2、随机去选出m个特征, m <<M,建立决策树
- 采取bootstrap抽样
为什么采用BootStrap抽样
- 为什么要随机抽样训练集?
- 如果不进行随机抽样,每棵树的训练集都一样,那么最终训练出的树分类结果也是完全一样的
- 为什么要有放回地抽样?
- 如果不是有放回的抽样,那么每棵树的训练样本都是不同的,都是没有交集的,这样每棵树都是“有偏的”,都是绝对“片面的”(当然这样说可能不对),也就是说每棵树训练出来都是有很大的差异的;而随机森林最后分类取决于多棵树(弱分类器)的投票表决。
API
-
class sklearn.ensemble.RandomForestClassifier(n_estimators=10, criterion=’gini’, max_depth=None, bootstrap=True, random_state=None, min_samples_split=2)
- 随机森林分类器
- n_estimators:integer,optional(default = 10)森林里的树木数量120,200,300,500,800,1200
- criteria:string,可选(default =“gini”)分割特征的测量方法
- max_depth:integer或None,可选(默认=无)树的最大深度 5,8,15,25,30
- max_features="auto”,每个决策树的最大特征数量
- If "auto", then
max_features=sqrt(n_features)
. - If "sqrt", then
max_features=sqrt(n_features)
(same as "auto"). - If "log2", then
max_features=log2(n_features)
. - If None, then
max_features=n_features
.
- If "auto", then
- bootstrap:boolean,optional(default = True)是否在构建树时使用放回抽样
- min_samples_split:节点划分最少样本数
- min_samples_leaf:叶子节点的最小样本数
- 超参数:n_estimator, max_depth, min_samples_split,min_samples_leaf
from sklearn.ensemble import RandomForestClassifier from sklearn.model_selection import GridSearchCV estimator = RandomForestClassifier() # 加入网格搜索与交叉验证 # 参数准备 param_dict = {"n_estimators": [120,200,300,500,800,1200], "max_depth": [5,8,15,25,30]} estimator = GridSearchCV(estimator, param_grid=param_dict, cv=3) estimator.fit(x_train, y_train) # 5)模型评估 # 方法1:直接比对真实值和预测值 y_predict = estimator.predict(x_test) print("y_predict:\n", y_predict) print("直接比对真实值和预测值:\n", y_test == y_predict) # 方法2:计算准确率 score = estimator.score(x_test, y_test) print("准确率为:\n", score) # 最佳参数:best_params_ print("最佳参数:\n", estimator.best_params_) # 最佳结果:best_score_ print("最佳结果:\n", estimator.best_score_) # 最佳估计器:best_estimator_ print("最佳估计器:\n", estimator.best_estimator_) # 交叉验证结果:cv_results_ print("交叉验证结果:\n", estimator.cv_results_)
结果:
y_predict: [0 0 0 0 1 1 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 1 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0 0 0 0 1 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 1 0 1 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 1 0 0 1 1 0 1 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 0 0 0 0 1 0 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 1 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 1 0 1 0 0 0 1 0 0 1 0 0 0 0 1 0 1 0 0 0 0 1 0 0 1 1 1 0 0 1 1 1 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 1 0 1 0 0 0 0 1] 直接比对真实值和预测值: 831 True 261 True 1210 True 1155 True 255 True 762 True 615 True 507 True 1175 True 301 True 1134 True 177 True 183 False 125 False 1093 True 1304 False 1124 True 798 False 1101 True 1239 False 1153 True 1068 False 846 True 148 True 478 True 642 True 1298 True 540 True 28 True 130 True ... 194 True 663 True 1209 True 117 False 595 False 1151 False 1143 True 1216 True 874 True 246 True 160 True 1208 True 682 True 307 True 67 True 961 True 400 True 923 False 866 True 134 True 613 True 242 True 320 False 829 True 94 True 1146 True 1125 False 386 True 1025 False 337 True Name: survived, Length: 329, dtype: bool 准确率为: 0.7872340425531915 最佳参数: {'max_depth': 5, 'n_estimators': 120} 最佳结果: 0.8363821138211383 最佳估计器: RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini', max_depth=5, max_features='auto', max_leaf_nodes=None, min_impurity_decrease=0.0, min_impurity_split=None, min_samples_leaf=1, min_samples_split=2, min_weight_fraction_leaf=0.0, n_estimators=120, n_jobs=None, oob_score=False, random_state=None, verbose=0, warm_start=False) 交叉验证结果: {'mean_fit_time': array([0.11182229, 0.19149677, 0.27871044, 0.4505314 , 0.72257209, 1.21950404, 0.15458934, 0.23542873, 0.37338622, 0.55880507, 0.90250031, 1.44036126, 0.13625924, 0.24566126, 0.39018901, 0.57973933, 0.94357061, 1.46748765, 0.15806643, 0.25924444, 0.3800021 , 0.60227998, 0.98656511, 1.5208021 , 0.15277807, 0.25416827, 0.37849299, 0.61238893, 1.00995 , 1.51009766]), 'std_fit_time': array([0.00438099, 0.00391445, 0.00445387, 0.00552127, 0.0178945 , 0.05956372, 0.00696462, 0.01180214, 0.01545986, 0.02345017, 0.01762821, 0.07661026, 0.00448709, 0.00753101, 0.01337304, 0.02401102, 0.02824846, 0.00723971, 0.00559061, 0.00539144, 0.03176938, 0.00900011, 0.0357836 , 0.02412509, 0.01049831, 0.00312499, 0.02043117, 0.03736237, 0.03896 , 0.01708367]), 'mean_score_time': array([0.01055225, 0.02124031, 0.02604191, 0.04676072, 0.06393997, 0.1221021 , 0.01392762, 0.02117666, 0.03027145, 0.04542494, 0.08080705, 0.11298935, 0.01059707, 0.02046402, 0.02975106, 0.04587412, 0.07316939, 0.14350526, 0.0142649 , 0.02011824, 0.02920715, 0.0444289 , 0.07418664, 0.11165055, 0.01248868, 0.02353628, 0.03232622, 0.04952399, 0.08569598, 0.11799375]), 'std_score_time': array([1.09863734e-03, 2.29822618e-03, 3.20843508e-03, 4.00866766e-03, 1.42997845e-03, 1.48818168e-02, 2.37098736e-03, 8.80449078e-04, 1.62827120e-03, 1.83137647e-03, 9.86835991e-03, 9.71738484e-03, 5.51943914e-04, 1.00782641e-03, 2.11610207e-03, 1.98464255e-03, 3.04582952e-03, 9.81828652e-03, 1.69302449e-03, 1.37694072e-03, 1.67724778e-03, 7.58986198e-05, 3.23449160e-03, 3.78348887e-03, 1.02684570e-03, 5.07326308e-03, 4.72586897e-03, 2.47344396e-03, 1.11438683e-02, 4.31988881e-03]), 'param_max_depth': masked_array(data=[5, 5, 5, 5, 5, 5, 8, 8, 8, 8, 8, 8, 15, 15, 15, 15, 15, 15, 25, 25, 25, 25, 25, 25, 30, 30, 30, 30, 30, 30], mask=[False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False], fill_value='?', dtype=object), 'param_n_estimators': masked_array(data=[120, 200, 300, 500, 800, 1200, 120, 200, 300, 500, 800, 1200, 120, 200, 300, 500, 800, 1200, 120, 200, 300, 500, 800, 1200, 120, 200, 300, 500, 800, 1200], mask=[False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False, False], fill_value='?', dtype=object), 'params': [{'max_depth': 5, 'n_estimators': 120}, {'max_depth': 5, 'n_estimators': 200}, {'max_depth': 5, 'n_estimators': 300}, {'max_depth': 5, 'n_estimators': 500}, {'max_depth': 5, 'n_estimators': 800}, {'max_depth': 5, 'n_estimators': 1200}, {'max_depth': 8, 'n_estimators': 120}, {'max_depth': 8, 'n_estimators': 200}, {'max_depth': 8, 'n_estimators': 300}, {'max_depth': 8, 'n_estimators': 500}, {'max_depth': 8, 'n_estimators': 800}, {'max_depth': 8, 'n_estimators': 1200}, {'max_depth': 15, 'n_estimators': 120}, {'max_depth': 15, 'n_estimators': 200}, {'max_depth': 15, 'n_estimators': 300}, {'max_depth': 15, 'n_estimators': 500}, {'max_depth': 15, 'n_estimators': 800}, {'max_depth': 15, 'n_estimators': 1200}, {'max_depth': 25, 'n_estimators': 120}, {'max_depth': 25, 'n_estimators': 200}, {'max_depth': 25, 'n_estimators': 300}, {'max_depth': 25, 'n_estimators': 500}, {'max_depth': 25, 'n_estimators': 800}, {'max_depth': 25, 'n_estimators': 1200}, {'max_depth': 30, 'n_estimators': 120}, {'max_depth': 30, 'n_estimators': 200}, {'max_depth': 30, 'n_estimators': 300}, {'max_depth': 30, 'n_estimators': 500}, {'max_depth': 30, 'n_estimators': 800}, {'max_depth': 30, 'n_estimators': 1200}], 'split0_test_score': array([0.82674772, 0.82066869, 0.82674772, 0.82674772, 0.82674772, 0.82674772, 0.80547112, 0.80851064, 0.80243161, 0.80243161, 0.80243161, 0.81155015, 0.79027356, 0.79635258, 0.79635258, 0.79331307, 0.79027356, 0.79027356, 0.79635258, 0.79027356, 0.79635258, 0.79331307, 0.79027356, 0.79331307, 0.79027356, 0.79635258, 0.79635258, 0.79331307, 0.7993921 , 0.79331307]), 'split1_test_score': array([0.85670732, 0.85670732, 0.85365854, 0.85365854, 0.85365854, 0.85365854, 0.85060976, 0.8597561 , 0.84756098, 0.85670732, 0.85670732, 0.85670732, 0.85365854, 0.85365854, 0.85365854, 0.8597561 , 0.85060976, 0.85670732, 0.85670732, 0.85365854, 0.85670732, 0.85365854, 0.85670732, 0.85060976, 0.85670732, 0.8597561 , 0.84756098, 0.85670732, 0.85060976, 0.85060976]), 'split2_test_score': array([0.82568807, 0.82262997, 0.82262997, 0.82568807, 0.82262997, 0.82262997, 0.80122324, 0.79510703, 0.80122324, 0.80428135, 0.80122324, 0.80122324, 0.80428135, 0.80122324, 0.80122324, 0.80122324, 0.80428135, 0.80428135, 0.80428135, 0.80733945, 0.80122324, 0.80122324, 0.80122324, 0.80122324, 0.79816514, 0.80428135, 0.80122324, 0.80428135, 0.80122324, 0.80122324]), 'mean_test_score': array([0.83638211, 0.83333333, 0.83434959, 0.83536585, 0.83434959, 0.83434959, 0.81910569, 0.82113821, 0.81707317, 0.82113821, 0.82012195, 0.82317073, 0.81605691, 0.81707317, 0.81707317, 0.81808943, 0.81504065, 0.81707317, 0.81910569, 0.81707317, 0.81808943, 0.81605691, 0.81605691, 0.81504065, 0.81504065, 0.82012195, 0.81504065, 0.81808943, 0.81707317, 0.81504065]), 'std_test_score': array([0.0143786 , 0.01654729, 0.01375658, 0.01294211, 0.01375658, 0.01375658, 0.02234414, 0.02784983, 0.02156378, 0.02516249, 0.02587446, 0.02408579, 0.02719639, 0.02594607, 0.02594607, 0.02963923, 0.02579309, 0.02860307, 0.02678467, 0.02679151, 0.02737927, 0.02678375, 0.02908969, 0.02535762, 0.0296384 , 0.02821188, 0.02308115, 0.02767166, 0.02372573, 0.02535762]), 'rank_test_score': array([ 1, 6, 3, 2, 3, 3, 12, 8, 17, 8, 10, 7, 23, 17, 17, 14, 26, 17, 12, 17, 14, 23, 23, 26, 26, 10, 26, 14, 17, 26], dtype=int32), 'split0_train_score': array([0.85801527, 0.85954198, 0.85648855, 0.85801527, 0.85648855, 0.85648855, 0.87633588, 0.87633588, 0.87633588, 0.87633588, 0.87633588, 0.87480916, 0.88244275, 0.88244275, 0.88244275, 0.88244275, 0.88244275, 0.88244275, 0.88244275, 0.88244275, 0.88244275, 0.88244275, 0.88244275, 0.88244275, 0.88244275, 0.88244275, 0.88244275, 0.88244275, 0.88244275, 0.88244275]), 'split1_train_score': array([0.84603659, 0.84756098, 0.84756098, 0.84756098, 0.84756098, 0.84756098, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049, 0.86128049]), 'split2_train_score': array([0.87214612, 0.87214612, 0.86757991, 0.87214612, 0.86757991, 0.86757991, 0.88736682, 0.88736682, 0.88584475, 0.88584475, 0.88584475, 0.88584475, 0.88736682, 0.88736682, 0.88736682, 0.88736682, 0.88736682, 0.88736682, 0.88736682, 0.88736682, 0.88736682, 0.88736682, 0.88736682, 0.88736682, 0.88736682, 0.88736682, 0.88736682, 0.88736682, 0.88736682, 0.88736682]), 'mean_train_score': array([0.85873266, 0.85974969, 0.85720981, 0.85924079, 0.85720981, 0.85720981, 0.87499439, 0.87499439, 0.87448704, 0.87448704, 0.87448704, 0.87397813, 0.87703002, 0.87703002, 0.87703002, 0.87703002, 0.87703002, 0.87703002, 0.87703002, 0.87703002, 0.87703002, 0.87703002, 0.87703002, 0.87703002, 0.87703002, 0.87703002, 0.87703002, 0.87703002, 0.87703002, 0.87703002]), 'std_train_score': array([0.01067124, 0.01003792, 0.00818859, 0.01007418, 0.00818859, 0.00818859, 0.01069186, 0.01069186, 0.01011317, 0.01011317, 0.01011317, 0.01004552, 0.01131658, 0.01131658, 0.01131658, 0.01131658, 0.01131658, 0.01131658, 0.01131658, 0.01131658, 0.01131658, 0.01131658, 0.01131658, 0.01131658, 0.01131658, 0.01131658, 0.01131658, 0.01131658, 0.01131658, 0.01131658])}