sklearn-docs/voting-classifier-decision-surface

1import numpy as np2import matplotlib.pyplot as plt3from matplotlib.colors import ListedColormap4from itertools import combinations5 6plt.rcParams['figure.dpi'] = 1007 8from sklearn.datasets import load_iris9 10from sklearn.tree import DecisionTreeClassifier11from sklearn.neighbors import KNeighborsClassifier12from sklearn.svm import SVC13from sklearn.ensemble import VotingClassifier14 15import gradio as gr16 17#==================================================18C1, C2, C3 = '#ff0000', '#ffff00', '#0000ff'19CMAP = ListedColormap([C1, C2, C3])20GRANULARITY = 0.0521 22FEATURE_NAMES = ["Sepal Length", "Sepal Width", "Petal Length", "Petal Width"]23TARGET_NAMES = ["Setosa", "Versicolour", "Virginica"]24MODEL_NAMES = ['DecisionTreeClassifier', 'KNeighborsClassifier', 'SupportVectorClassifier', 'VotingClassifier']25 26iris = load_iris()27#==================================================28def get_decision_surface(X, y, model):29    x_min, x_max = X[:, 0].min() - 1, X[:, 0].max() + 130    y_min, y_max = X[:, 1].min() - 1, X[:, 1].max() + 131    xrange = np.arange(x_min, x_max, GRANULARITY)32    yrange = np.arange(y_min, y_max, GRANULARITY)33    xx, yy = np.meshgrid(xrange, yrange)34 35    Z = model.predict(np.c_[xx.ravel(), yy.ravel()])36    Z = Z.reshape(xx.shape)37 38    return xx, yy, Z39 40def create_plot(feature_string, max_depth, n_neighbors, gamma, weight1, weight2, weight3):41 42    feature_list = feature_string.split(',')43    feature_list = [s.strip() for s in feature_list]44    idx_x = FEATURE_NAMES.index(feature_list[0])45    idx_y = FEATURE_NAMES.index(feature_list[1])46 47    X = iris.data[:, [idx_x, idx_y]]48    y = iris.target49 50    rnd_idx = np.random.permutation(X.shape[0])51    X = X[rnd_idx]52    y = y[rnd_idx]53 54    clf1 = DecisionTreeClassifier(max_depth=max_depth)55    clf2 = KNeighborsClassifier(n_neighbors=n_neighbors, n_jobs=-1)56    clf3 = SVC(gamma=gamma, kernel="rbf", probability=True)57    eclf = VotingClassifier(58        estimators=[("dt", clf1), ("knn", clf2), ("svc", clf3)],59        voting="soft",60        weights=[weight1, weight2, weight3],61    )62 63    clf1.fit(X, y)64    clf2.fit(X, y)65    clf3.fit(X, y)66    eclf.fit(X, y)67 68    fig, _ = plt.subplots(2, 2, figsize=(7, 7), sharex=True, sharey=True)69 70    for i, clf in enumerate([clf1, clf2, clf3, eclf]):71        xx, yy, Z = get_decision_surface(X, y, clf)72 73        ax = fig.add_subplot(2, 2, i+1)74        ax.set_axis_off()75        ax.contourf(xx, yy, Z, cmap=CMAP, alpha=0.65)76 77        for j, label in enumerate(TARGET_NAMES):78            X_label = X[y==j,:]79            y_label = y[y==j]80            ax.scatter(X_label[:, 0], X_label[:, 1], c=[[C1], [C2], [C3]][j]*len(y_label), edgecolor='k', s=40, label=label)81        82        ax.legend()83        ax.set_title(f'{MODEL_NAMES[i]}')84        85    fig.supxlabel(feature_list[0]); fig.supylabel(feature_list[1])86    fig.set_tight_layout(True)87    fig.set_constrained_layout(True)88    return fig89 90info = '''91# Voting Classifier Decision Surface92 93This app plots the decision surface of four classifiers on two selected features of the Iris dataset: DecisionTreeClassifier, KNeighborsClassifier, SupportVectorClassifier, and a VotingClassifier from all of them.94 95Use the controls below to tune the parameters of the classifiers and the weights of each of them in the soft voting classifier and click submit. The more weight you assign to a classifier, the more importance will be assigned to its predictions compared to the other classifiers in the vote.96 97Created by [@huabdul](https://huggingface.co/huabdul) based on [scikit-learn docs](https://scikit-learn.org/stable/auto_examples/ensemble/plot_voting_decision_regions.html).98'''99 100with gr.Blocks(analytics_enabled=False) as demo:101    selections = combinations(FEATURE_NAMES, 2)102    selections = [f'{s[0]}, {s[1]}' for s in selections]103 104    with gr.Row():105        with gr.Column():106            gr.Markdown(info)107            dd = gr.Dropdown(selections, value=selections[0], interactive=True, label="Input features")108            with gr.Row(): 109                with gr.Column(min_width=100):110                    slider_max_depth = gr.Slider(1, 50, value=4, step=1, label='max_depth (DecisionTree)')111                    slider_n_neighbors = gr.Slider(1, 20, value=7, step=1, label='n_neighbors (KNN)')112                    slider_gamma = gr.Slider(0, 10, value=0.1, step=0.1, label='gamma (SVC)')113                with gr.Column(min_width=100):114                    slider_w1 = gr.Slider(0, 10, value=2, step=0.1, label='DecisionTreeClassifier weight')115                    slider_w2 = gr.Slider(0, 10, value=1, step=0.1, label='KNeighborsClassifier weight')116                    slider_w3 = gr.Slider(0, 10, value=2, step=0.1, label='SVC weight')117    118            btn = gr.Button(value='Submit')119        120        with gr.Column():121            plot = gr.Plot(show_label=False)122 123    btn.click(create_plot, inputs=[dd, slider_max_depth, slider_n_neighbors, slider_gamma, slider_w1, slider_w2, slider_w3], outputs=[plot])124    demo.load(create_plot, inputs=[dd, slider_max_depth, slider_n_neighbors, slider_gamma, slider_w1, slider_w2, slider_w3], outputs=[plot])125 126demo.launch()127#==================================================