PCA/PCA.py at main · AreaProgram/PCA · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib.patches as mpatches
import numpy as np
from sklearn.preprocessing import StandardScaler
df = pd.read_csv("/Users/arya/Downloads/cont_var_91_del.csv", header = None)

color = pd.read_csv("/Users/arya/Downloads/cont_var_91_del_labels1.csv", header = None).iloc[:,25].values

numbers = df.iloc[:, 25].values

X = df.iloc[:, :25].values

'''
PCA Calculation Functions

'''
def standardize_eigen(X):
    mean_vec = np.mean(X, axis=0)
    X_centered = X - mean_vec
    cov = np.cov(X_centered.T)
    eig_val, eig_vec = np.linalg.eig(cov)
    sorted_idx = np.argsort(eig_val)[::-1]
    return eig_val[sorted_idx], eig_vec[:, sorted_idx], X_centered

def explained_variance(eig_val):
    total = np.sum(eig_val)
    return total, np.cumsum(eig_val / total)

def pca_transform(X_centered, eig_vec, n_components=2):
    return np.dot(X_centered, eig_vec[:, :n_components])

'''
PCA Function
'''


def pca(X,components,Y,colors, specific_number = None):
    eig_val, eig_vec, X_centered = standardize_eigen(X)
    X_pca = pca_transform(X_centered, eig_vec, n_components=components)
    color = ""


    #Creates the plot figure; you may have to configure these dimensions to your display screen.
    plt.figure(figsize=(15, 10))

    # This is the color dictionary that is predefined with the links between the labels and the colors
    colordict = {"Omi+": "blue", "Omi+ ?":"brown", "ENJ":"red", "Omi- ?":"green", "Omi-":"purple"}
    legend_array = []

    # This is how the colors are designated before plotting the points, to ensure that the labels are mapped to the right color.
    if type(colors) != str:
        for i, label in enumerate(colors):
            color = colordict[colors[i]]
            plt.scatter(X_pca[i, components - 2],
                            X_pca[i, components - 1], # components - 2 would refer to PC1, components - 1 is PC2 (only if components is 2)
                            c = color,
                            label=label,
                            s=20)

    else:
        color = colors
        plt.scatter(X_pca[i, components - 2],
                    X_pca[i, components - 1], # components - 2 would refer to PC1, components - 1 is PC2 (only if components is 2)
                    c = color,
                    label=label,
                    s=20)
    #Checks if you have defined a specific number to be labeled, otherwise it will label all the points
    if specific_number != None:
        for j in specific_number:
            i = list(Y).index(j)
            plt.text(X_pca[i, components - 2],
                    X_pca[i, components - 1],
                    str(Y[i]),
                    fontsize=8,
                    alpha=0.6)
    else:
        for i in range(len(X_pca)):
            plt.text(X_pca[i, components - 2],
                    X_pca[i, components - 1],
                    str(Y[i]),
                    fontsize=8,
                    alpha=0.6)

    # Defines the legend by using the labels and colors
    for i, j in colordict.items():
        legend_array.append(mpatches.Patch(color = j, label = i))
    plt.xlabel(f"Principal Component {components - 1}")
    plt.ylabel(f"Principal Component {components}")
    plt.title(f"PCA - Components {components - 1} vs {components}")
    plt.grid(True)
    plt.tight_layout()
    plt.legend(title = "Group", handles = legend_array)
    plt.show()
# This is an optional scaler that scales the numbers to a specific range. It does not affect the clustering of the points, only the axis values.
X_scaled = StandardScaler().fit_transform(X)

pca(X_scaled,2, numbers, colors = color, specific_number=[13])