如何绘制GridSearchCV的网格分数？

Question

我正在寻找一种在sklearn中从GridSearchCV图形化grid_scores_的方法。在这个例子中，我试图网格搜索SVR算法的最佳gamma和C参数。我的代码如下：

    C_range = 10.0 ** np.arange(-4, 4)
    gamma_range = 10.0 ** np.arange(-4, 4)
    param_grid = dict(gamma=gamma_range.tolist(), C=C_range.tolist())
    grid = GridSearchCV(SVR(kernel='rbf', gamma=0.1),param_grid, cv=5)
    grid.fit(X_train,y_train)
    print(grid.grid_scores_)

运行代码并打印网格分数后，我得到以下结果：

[mean: -3.28593, std: 1.69134, params: {'gamma': 0.0001, 'C': 0.0001}, mean: -3.29370, std: 1.69346, params: {'gamma': 0.001, 'C': 0.0001}, mean: -3.28933, std: 1.69104, params: {'gamma': 0.01, 'C': 0.0001}, mean: -3.28925, std: 1.69106, params: {'gamma': 0.1, 'C': 0.0001}, mean: -3.28925, std: 1.69106, params: {'gamma': 1.0, 'C': 0.0001}, mean: -3.28925, std: 1.69106, params: {'gamma': 10.0, 'C': 0.0001},etc] 

我想根据gamma和C参数可视化所有分数（平均值）。我想要获得的图表应如下所示：

其中x轴是伽马，y轴是平均分数（在这种情况下是均方根误差），不同的线代表不同的C值。

Answer 1

@sascha显示的代码是正确的。但是，grid_scores_属性很快就会被弃用。最好使用cv_results属性。

它可以以类似@sascha方法的方式实现：

def plot_grid_search(cv_results, grid_param_1, grid_param_2, name_param_1, name_param_2):
    # Get Test Scores Mean and std for each grid search
    scores_mean = cv_results['mean_test_score']
    scores_mean = np.array(scores_mean).reshape(len(grid_param_2),len(grid_param_1))

    scores_sd = cv_results['std_test_score']
    scores_sd = np.array(scores_sd).reshape(len(grid_param_2),len(grid_param_1))

    # Plot Grid search scores
    _, ax = plt.subplots(1,1)

    # Param1 is the X-axis, Param 2 is represented as a different curve (color line)
    for idx, val in enumerate(grid_param_2):
        ax.plot(grid_param_1, scores_mean[idx,:], '-o', label= name_param_2 + ': ' + str(val))

    ax.set_title("Grid Search Scores", fontsize=20, fontweight='bold')
    ax.set_xlabel(name_param_1, fontsize=16)
    ax.set_ylabel('CV Average Score', fontsize=16)
    ax.legend(loc="best", fontsize=15)
    ax.grid('on')

# Calling Method 
plot_grid_search(pipe_grid.cv_results_, n_estimators, max_features, 'N Estimators', 'Max Features')

以上结果如下：

Answer 2

from sklearn.svm import SVC
from sklearn.grid_search import GridSearchCV
from sklearn import datasets
import matplotlib.pyplot as plt
import seaborn as sns
import numpy as np

digits = datasets.load_digits()
X = digits.data
y = digits.target

clf_ = SVC(kernel='rbf')
Cs = [1, 10, 100, 1000]
Gammas = [1e-3, 1e-4]
clf = GridSearchCV(clf_,
            dict(C=Cs,
                 gamma=Gammas),
                 cv=2,
                 pre_dispatch='1*n_jobs',
                 n_jobs=1)

clf.fit(X, y)

scores = [x[1] for x in clf.grid_scores_]
scores = np.array(scores).reshape(len(Cs), len(Gammas))

for ind, i in enumerate(Cs):
    plt.plot(Gammas, scores[ind], label='C: ' + str(i))
plt.legend()
plt.xlabel('Gamma')
plt.ylabel('Mean score')
plt.show()

• 代码基于this。
• 只有令人费解的部分：sklearn会始终尊重C＆amp; C的顺序伽玛 - ＆gt;官方示例使用此“订购”

输出：

Answer 3

我想做类似的事情（但可以扩展到大量参数），这是我生成输出群图的解决方案：

score = pd.DataFrame(gs_clf.grid_scores_).sort_values(by='mean_validation_score', ascending = False)
for i in parameters.keys():
    print(i, len(parameters[i]), parameters[i])
score[i] = score.parameters.apply(lambda x: x[i])
l =['mean_validation_score'] + list(parameters.keys())
for i in list(parameters.keys()):
    sns.swarmplot(data = score[l], x = i, y = 'mean_validation_score')
    #plt.savefig('170705_sgd_optimisation//'+i+'.jpg', dpi = 100)
    plt.show()

Answer 4

遍历参数网格的顺序是确定性的，这样可以对其进行重新整形和直接绘制。像这样：

backdoor

Answer 5

这是一个利用seaborn pointplot的解决方案。这种方法的优点是，当您搜索两个以上的参数时，它将允许您绘制结果

import seaborn as sns
import pandas as pd

def plot_cv_results(cv_results, param_x, param_z, metric='mean_test_score'):
    """
    cv_results - cv_results_ attribute of a GridSearchCV instance (or similar)
    param_x - name of grid search parameter to plot on x axis
    param_z - name of grid search parameter to plot by line color
    """
    cv_results = pd.DataFrame(cv_results)
    col_x = 'param_' + param_x
    col_z = 'param_' + param_z
    fig, ax = plt.subplots(1, 1, figsize=(11, 8))
    sns.pointplot(x=col_x, y=metric, hue=col_z, data=cv_results, ci=99, n_boot=64, ax=ax)
    ax.set_title("CV Grid Search Results")
    ax.set_xlabel(param_x)
    ax.set_ylabel(metric)
    ax.legend(title=param_z)
    return fig

xgboost的用法示例：

from xgboost import XGBRegressor
from sklearn import GridSearchCV

params = {
    'max_depth': [3, 6, 9, 12], 
    'gamma': [0, 1, 10, 20, 100],
    'min_child_weight': [1, 4, 16, 64, 256],
}
model = XGBRegressor()
grid = GridSearchCV(model, params, scoring='neg_mean_squared_error')
grid.fit(...)
fig = plot_cv_results(grid.cv_results_, 'gamma', 'min_child_weight')

这将产生一个图形，该图形在x轴上显示gamma正则化参数，在线条颜色中显示min_child_weight正则化参数，以及任何其他网格搜索参数（在这种情况下为{{1 }}）将通过seaborn点图的99％置信区间的分布来描述。

*请注意，在下面的示例中，我从上面的代码中稍微更改了外观。

Answer 6

我在xgboost上使用了具有不同学习率，最大深度和估计量的网格搜索。

gs_param_grid = {'max_depth': [3,4,5], 
                 'n_estimators' : [x for x in range(3000,5000,250)],
                 'learning_rate':[0.01,0.03,0.1]
                }
gbm = XGBRegressor()
grid_gbm = GridSearchCV(estimator=gbm, 
                        param_grid=gs_param_grid, 
                        scoring='neg_mean_squared_error', 
                        cv=4, 
                        verbose=1
                       )
grid_gbm.fit(X_train,y_train)

要使用不同的学习率来创建误差与估计量之间的关系图，我使用了以下方法：

y=[]
cvres = grid_gbm.cv_results_
best_md=grid_gbm.best_params_['max_depth']
la=gs_param_grid['learning_rate']
n_estimators=gs_param_grid['n_estimators']

for mean_score, params in zip(cvres["mean_test_score"], cvres["params"]):
    if params["max_depth"]==best_md:
        y.append(np.sqrt(-mean_score))


y=np.array(y).reshape(len(la),len(n_estimators))

%matplotlib inline
plt.figure(figsize=(8,8))
for y_arr, label in zip(y, la):
    plt.plot(n_estimators, y_arr, label=label)

plt.title('Error for different learning rates(keeping max_depth=%d(best_param))'%best_md)
plt.legend()
plt.xlabel('n_estimators')
plt.ylabel('Error')
plt.show()

可以在此处查看该图： Result

请注意，可以类似地针对误差与估计器数量（具有不同的最大深度）（或根据用户情况的任何其他参数）创建图形。

Answer 7

当我尝试绘制平均得分与否时，这对我有用。随机森林中的树木。 reshape（）函数有助于找出平均值。

param_n_estimators = cv_results['param_n_estimators']
param_n_estimators = np.array(param_n_estimators)
mean_n_estimators = np.mean(param_n_estimators.reshape(-1,5), axis=0)

mean_test_scores = cv_results['mean_test_score']
mean_test_scores = np.array(mean_test_scores)
mean_test_scores = np.mean(mean_test_scores.reshape(-1,5), axis=0)

mean_train_scores = cv_results['mean_train_score']
mean_train_scores = np.array(mean_train_scores)
mean_train_scores = np.mean(mean_train_scores.reshape(-1,5), axis=0)

Answer 8

为了在调整多个超参数时绘制结果，我所做的是将所有参数固定为它们的最佳值（一个参数除外），并为每个参数绘制另一个参数的平均得分。

def plot_search_results(grid):
    """
    Params: 
        grid: A trained GridSearchCV object.
    """
    ## Results from grid search
    results = grid.cv_results_
    means_test = results['mean_test_score']
    stds_test = results['std_test_score']
    means_train = results['mean_train_score']
    stds_train = results['std_train_score']

    ## Getting indexes of values per hyper-parameter
    masks=[]
    masks_names= list(grid.best_params_.keys())
    for p_k, p_v in grid.best_params_.items():
        masks.append(list(results['param_'+p_k].data==p_v))

    params=grid.param_grid

    ## Ploting results
    fig, ax = plt.subplots(1,len(params),sharex='none', sharey='all',figsize=(20,5))
    fig.suptitle('Score per parameter')
    fig.text(0.04, 0.5, 'MEAN SCORE', va='center', rotation='vertical')
    pram_preformace_in_best = {}
    for i, p in enumerate(masks_names):
        m = np.stack(masks[:i] + masks[i+1:])
        pram_preformace_in_best
        best_parms_mask = m.all(axis=0)
        best_index = np.where(best_parms_mask)[0]
        x = np.array(params[p])
        y_1 = np.array(means_test[best_index])
        e_1 = np.array(stds_test[best_index])
        y_2 = np.array(means_train[best_index])
        e_2 = np.array(stds_train[best_index])
        ax[i].errorbar(x, y_1, e_1, linestyle='--', marker='o', label='train')
        ax[i].errorbar(x, y_2, e_2, linestyle='-', marker='^',label='test' )
        ax[i].set_xlabel(p.upper())

    plt.show()

Result

Answer 9

这里有完整的代码，可以生成绘图，因此您可以使用GridSearchCV完全可视化多达3个参数的变化。这是运行代码时将看到的内容：

1. 参数1（x轴）
2. 交叉验证平均得分（y轴）
3. Parameter2（为每个不同的Parameter2值绘制的额外线条，带有图例供参考）
4. Parameter3（将为每个不同的Parameter3值弹出额外的图表，使您可以查看这些不同图表之间的差异）

对于每条绘制的线，还显示了基于您正在运行的多个CV所期望的交叉验证平均得分的标准偏差。享受吧！

from sklearn import tree
from sklearn import model_selection
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from sklearn.preprocessing import LabelEncoder
from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn.datasets import load_digits

digits = load_digits()
X, y = digits.data, digits.target

Algo = [['DecisionTreeClassifier', tree.DecisionTreeClassifier(),  # algorithm
             'max_depth', [1, 2, 4, 6, 8, 10, 12, 14, 18, 20, 22, 24, 26, 28, 30],  # Parameter1
             'max_features', ['sqrt', 'log2', None],  # Parameter2
                 'criterion', ['gini', 'entropy']]]  # Parameter3


def plot_grid_search(cv_results, grid_param_1, grid_param_2, name_param_1, name_param_2, title):
    # Get Test Scores Mean and std for each grid search

    grid_param_1 = list(str(e) for e in grid_param_1)
    grid_param_2 = list(str(e) for e in grid_param_2)
    scores_mean = cv_results['mean_test_score']
    scores_std = cv_results['std_test_score']
    params_set = cv_results['params']

    scores_organized = {}
    std_organized = {}
    std_upper = {}
    std_lower = {}
    for p2 in grid_param_2:
        scores_organized[p2] = []
        std_organized[p2] = []
        std_upper[p2] = []
        std_lower[p2] = []
        for p1 in grid_param_1:
            for i in range(len(params_set)):
                if str(params_set[i][name_param_1]) == str(p1) and str(params_set[i][name_param_2]) == str(p2):
                    mean = scores_mean[i]
                    std = scores_std[i]
                    scores_organized[p2].append(mean)
                    std_organized[p2].append(std)
                    std_upper[p2].append(mean + std)
                    std_lower[p2].append(mean - std)

    _, ax = plt.subplots(1, 1)

    # Param1 is the X-axis, Param 2 is represented as a different curve (color line)
    # plot means
    for key in scores_organized.keys():
        ax.plot(grid_param_1, scores_organized[key], '-o', label= name_param_2 + ': ' + str(key))
        ax.fill_between(grid_param_1, std_lower[key], std_upper[key], alpha=0.1)

    ax.set_title(title)
    ax.set_xlabel(name_param_1)
    ax.set_ylabel('CV Average Score')
    ax.legend(loc="best")
    ax.grid('on')
    plt.show()

dataset = 'Titanic'

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
cv_split = model_selection.KFold(n_splits=10, random_state=2)



for i in range(len(Algo)):

    name = Algo[0][0]
    alg = Algo[0][1]
    param_1_name = Algo[0][2]
    param_1_range = Algo[0][3]
    param_2_name = Algo[0][4]
    param_2_range = Algo[0][5]
    param_3_name = Algo[0][6]
    param_3_range = Algo[0][7]

    for p in param_3_range:
        # grid search
        param = {
            param_1_name: param_1_range,
            param_2_name: param_2_range,
            param_3_name: [p]
        }
        grid_test = GridSearchCV(alg, param_grid=param, scoring='accuracy', cv=cv_split)
        grid_test.fit(X_train, y_train)
        plot_grid_search(grid_test.cv_results_, param[param_1_name], param[param_2_name], param_1_name, param_2_name, dataset + ' GridSearch Scores: ' + name + ', ' + param_3_name + '=' + str(p))

    param = {
        param_1_name: param_1_range,
        param_2_name: param_2_range,
        param_3_name: param_3_range
    }
    grid_final = GridSearchCV(alg, param_grid=param, scoring='accuracy', cv=cv_split)
    grid_final.fit(X_train, y_train)
    best_params = grid_final.best_params_
    alg.set_params(**best_params)

Answer 10

@nathandrake尝试以下基于@ david-alvarez的代码改编而成的内容：

def plot_grid_search(cv_results, metric, grid_param_1, grid_param_2, name_param_1, name_param_2):
    # Get Test Scores Mean and std for each grid search
    scores_mean = cv_results[('mean_test_' + metric)]
    scores_sd = cv_results[('std_test_' + metric)]

    if grid_param_2 is not None:
        scores_mean = np.array(scores_mean).reshape(len(grid_param_2),len(grid_param_1))
        scores_sd = np.array(scores_sd).reshape(len(grid_param_2),len(grid_param_1))

    # Set plot style
    plt.style.use('seaborn')

    # Plot Grid search scores
    _, ax = plt.subplots(1,1)

    if grid_param_2 is not None:
        # Param1 is the X-axis, Param 2 is represented as a different curve (color line)
        for idx, val in enumerate(grid_param_2):
            ax.plot(grid_param_1, scores_mean[idx,:], '-o', label= name_param_2 + ': ' + str(val))
    else:
        # If only one Param1 is given
        ax.plot(grid_param_1, scores_mean, '-o')

    ax.set_title("Grid Search", fontsize=20, fontweight='normal')
    ax.set_xlabel(name_param_1, fontsize=16)
    ax.set_ylabel('CV Average ' + str.capitalize(metric), fontsize=16)
    ax.legend(loc="best", fontsize=15)
    ax.grid('on')

如您所见，我添加了支持包括多个指标的网格搜索的功能。您只需在对绘图功能的调用中指定要绘制的指标即可。

此外，如果您的网格搜索仅调整了一个参数，则只需为grid_param_2和name_param_2指定None。

按如下方式调用它：

plot_grid_search(grid_search.cv_results_,
                 'Accuracy',
                 list(np.linspace(0.001, 10, 50)), 
                 ['linear', 'rbf'],
                 'C',
                 'kernel')