LDA com iris

 import numpy as np

class LDA:

    def __init__(self, n_components):

        self.n_components = n_components

        self.linear_discriminants = None

    def fit(self, x, y):

        n_features = x.shape[1]

        class_labels = np.unique(y)

        mean_overall = np.mean(x, axis=0)

        S_W = np.zeros((n_features,n_features))

        S_B = np.zeros((n_features,n_features))

        for c in class_labels:

            X_c = x[y==c]

            mean_c = np.mean(X_c, axis=0)

            S_W+= (X_c - mean_c).T.dot(X_c-mean_c)

            n_c = X_c.shape[0]

            mean_diff = (mean_c - mean_overall).reshape(n_features,1)

            S_B += n_c*(mean_diff).dot(mean_diff.T)

        A = np.linalg.inv(S_W).dot(S_B)

        eigenvalues, eigenvectors = np.linalg.eig(A)

        eigenvectors = eigenvectors.T

        idxs= np.argsort(abs(eigenvalues))[::-1]

        eigenvalues = eigenvalues[idxs]

        eigenvectors = eigenvectors[idxs]

        self.linear_discriminants = eigenvectors[0:self.n_components]

        

    def transform (self, x):

        return np.dot(x, self.linear_discriminants.T)

#### pra ser igual do PCA

from sklearn import datasets

#pip install lda

import matplotlib.pyplot as plt

#import numpy as np

#from lda import LDA

from sklearn.discriminant_analysis import LinearDiscriminantAnalysis as LDA

data = datasets.load_iris()

x=data.data

y=data.target

lda= LDA(2) #(dois exios)

lda.fit (x,y)

X_projected = lda.transform(x)

print ('shape of x', x.shape)

print ('shape of transformed x:', X_projected.shape)

x1 = X_projected[:,0]

x2 = X_projected[:,1]

plt.scatter (x1,x2,c=y, edgecolor='none', alpha=0.8, cmap=plt.cm.get_cmap('viridis',3))

plt.xlabel('linear discriminant 1')

plt.ylabel('linear discriminant 2')

plt.colorbar()

plt.show()