ds3l4q3corrected

'''Name: Ritam Chakraborty
   Roll: B20127
   Phone: 7439257709'''

# import libraries
import pandas as pd
import numpy as np
from numpy.linalg import det, inv
# for computing confusion matrix and accuracy score
from sklearn.metrics import confusion_matrix, accuracy_score
# for making comparator to sort the distances and corresponding classes based on distances

def prob(x, inv_cov_mat, mu_mat, Pci):
    diff = x-mu_mat
    diff_T = diff.T
    dotprod = np.dot(diff.T, inv_cov_mat)
    prod = np.dot(dotprod, diff)
    return np.log(Pci)+0.5*np.log(det(inv_cov_mat))-11.5*np.log(2*np.pi)-0.5*prod

df_train = pd.read_csv("SteelPlateFaults-train.csv")
df_test = pd.read_csv("SteelPlateFaults-test.csv")
# remove the following attributes since they make the covariance matrix singular
delete_items = ["TypeOfSteel_A300", "TypeOfSteel_A400", "X_Minimum", "Y_Minimum"]
for items in delete_items:
    df_train.__delitem__(items)
    df_test.__delitem__(items)

# separate the classes from the training dataset
df_class1 = df_train[df_train["Class"] == 1].copy()
df_class1.__delitem__("Class")
df_class0 = df_train[df_train["Class"] == 0].copy()
df_class0.__delitem__("Class")

# compute mean vectors from both classes
mu1 = df_class1.mean().to_numpy()
mu0 = df_class0.mean().to_numpy()
inv1 = inv(df_class1.cov())
inv0 = inv(df_class0.cov())
pd.DataFrame(mu0).to_csv("mean0.csv")
pd.DataFrame(mu1).to_csv("mean1.csv")
df_class1.cov().to_csv("cov1")
df_class0.cov().to_csv("cov0")

# predict based on bayes classifier
pred = []
for ind in df_test.index:
    p1 = prob(df_test.iloc[ind, :23].to_numpy(), inv1, mu1, 509/(509+273))
    p0 = prob(df_test.iloc[ind, :23].to_numpy(), inv0, mu0, 273/(509+273))
    if p1 > p0:
        pred.append(1)
    else:
        pred.append(0)

#print the confusion matrix and accuracy scores
print("Confusion matrix:")
print(confusion_matrix(df_test["Class"], pred))
print("Accuracy score:")
print(accuracy_score(df_test["Class"], pred))