SongLynn 2019-07-01
分裂规则
减少过拟合
样本均衡问题
回归问题
随机性
样本均衡问题
需要注意的是权重对于bootstrap的使用并没有影响,即bootstrap方法始终是等概率地从N个样本中选择,sklearn中的源码如下
if forest.bootstrap: n_samples = X.shape[0] if sample_weight is None: curr_sample_weight = np.ones((n_samples,), dtype=np.float64) else: curr_sample_weight = sample_weight.copy() #已经包含了class_weight设为'balanced'或dict类型时的类别权重 indices = _generate_sample_indices(tree.random_state, n_samples) #bootstrap sample_counts = np.bincount(indices, minlength=n_samples) curr_sample_weight *= sample_counts #根据新的样本集合中每个原始样本的个数来调整样本权重 ### 根据类别权重调整样本权重 if class_weight == 'subsample': with catch_warnings(): simplefilter('ignore', DeprecationWarning) curr_sample_weight *= compute_sample_weight('auto', y, indices) elif class_weight == 'balanced_subsample': curr_sample_weight *= compute_sample_weight('balanced', y, indices) tree.fit(X, y, sample_weight=curr_sample_weight, check_input=False) else: tree.fit(X, y, sample_weight=sample_weight, check_input=False)
OOB(out-of-bag estimate)
特征重要性