LinearDiscriminantAnalysis.py

"""
Get coefficients from Linear Discriminant Analysis.
Written by: Lincoln Huber
"""

from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
import pandas as pd
import numpy as np
import math
import sys
import os

# get data file
path = 'iris_dataset.csv'

if os.path.exists(path):
    # create dataframe from file
    df = pd.read_csv(path)

    # get data and labels
    labels = df['class'].values.reshape(-1, 1)
    data = df.drop(['class'], axis=1)

    # get LDA
    lda = LinearDiscriminantAnalysis(n_components=1)

    # train
    lda.fit(data, labels.ravel())

    # get weights
    angles = lda.coef_[0].copy()

    # normalize
    angles = angles / max(abs(angles))

    # get degrees
    for i in range(len(angles)):
        angles[i] = np.arccos(angles[i])
        angles[i] = math.degrees(angles[i])

    # inform DV program of angles
    for i in range(len(angles)):
        print(angles[i])

    # normalize weights
    weights = lda.coef_[0].copy()
    weights = weights / max(abs(weights))

    # apply weights to class means
    weightedMeans = lda.means_ * weights

    # get sum of feature values for each class
    weightedMeanSum1 = sum(weightedMeans[0])
    weightedMeanSum2 = sum(weightedMeans[1])

    # add weighted class sums then divide by 2 to get threshold
    print((weightedMeanSum1 + weightedMeanSum2) / 2)