from sklearn import datasets
from sklearn.decomposition import PCA
from sklearn.preprocessing import scale
from pandas import DataFrame
iris = datasets.load_iris()
iris = data1 = DataFrame(data= np.c_[iris['data'], iris['target']],
                            columns= iris['feature_names'] + ['class'])

def plot_pca(data,classes, classes_label=None):
    fig, (axe_2d,axe_info) = plt.subplots(1,2)
    data_scaled = scale(data)
    classes_unique = np.unique(classes)
    if classes_label is None:
        classes_label = ["Class %i" % i for i in classes_unique]
    n_axis = np.shape(data)[1]
    pca = PCA()
    x_reduced = pca.fit_transform(data_scaled)
    var = pca.explained_variance_ratio_
    cumul_prop = np.cumsum(var)

    # Plot 2D
    for class_value, class_label in zip(classes_unique,classes_label):
        selector = classes==class_value
        axe_2d.scatter(x_reduced[selector,0],x_reduced[selector,1], label=class_label)
    axe_2d.set_title("2D plot")
    axe_2d.set_xlabel("Component 1")
    axe_2d.set_xlabel("Component 2")
    axe_2d.legend()
    # Plot bar
    axe_info.bar(range(n_axis),cumul_prop)
    axe_info.set_title("Components cumulative proportion")
    axe_info.set_xlabel("Components")
    axe_info.set_xlabel("Cumulative proportion")

    for axe in (axe_2d, axe_info):
        axe.grid()

plot_pca(iris.values[:,:-1],iris.values[:,-1])
plt.show()