import matplotlib.pyplot as plt

0.2*0.95+0.8*0.32

0.2*0.95/0.85

(0.2235294117647059 - 0.2*0.95)/0.8

2*(1/5)+3*(1/5)+4*(1/5)+5*(1/5)+6*(1/5)

x=[0]*20 y=[0]*20 for i in range(20): x[i]=i+1 y[i]=4*(i+1)*(5/6)**(i+1) plt.scatter(x,y)

y[5]

y[6]

import numpy as np def load_data(): with np.load('enron.npz') as data: return dict(**data) data = load_data() trainFeat = data['trainFeat'] trainLabels = data['trainLabels'] testFeat = data['testFeat'] testLabels = data['testLabels'] vocab = data['vocab'] W = len(vocab)

vocabInds = np.arange(W) trainHam = trainFeat[trainLabels == 0][:, vocabInds] trainSpam = trainFeat[trainLabels == 1][:, vocabInds] numHam = len(trainHam) numSpam = len(trainSpam) countsHam = np.sum(trainHam, axis=0) countsSpam = np.sum(trainSpam, axis=0) p_11_=countsSpam/numSpam p_10_=countsHam/numHam p_01_=1-p_11_ p_00_=1-p_10_ print(p_11_) print(p_10_) print(p_01_) print(p_00_)

vocabInds=179 trainHam = trainFeat[trainLabels == 0][:, vocabInds] trainSpam = trainFeat[trainLabels == 1][:, vocabInds] numHam = len(trainHam) numSpam = len(trainSpam) countsHam = np.sum(trainHam, axis=0) countsSpam = np.sum(trainSpam, axis=0) p_11=countsSpam/numSpam p_10=countsHam/numHam print(p_11) print(p_10)

predictLables=testFeat[:,vocabInds] matchedPrediction= predictLables == testLabels accuracy= np.sum(matchedPrediction)/len(testLabels) print(accuracy)

vocabInds=859 trainHam = trainFeat[trainLabels == 0][:, vocabInds] trainSpam = trainFeat[trainLabels == 1][:, vocabInds] numHam = len(trainHam) numSpam = len(trainSpam) countsHam = np.sum(trainHam, axis=0) countsSpam = np.sum(trainSpam, axis=0) p_11=countsSpam/numSpam p_10=countsHam/numHam print(p_11) print(p_10)

predictLables=testFeat[:,vocabInds] matchedPrediction= predictLables == testLabels accuracy= 1-(np.sum(matchedPrediction)/len(testLabels)) print(accuracy)

vocabInds=2211 trainHam = trainFeat[trainLabels == 0][:, vocabInds] trainSpam = trainFeat[trainLabels == 1][:, vocabInds] numHam = len(trainHam) numSpam = len(trainSpam) countsHam = np.sum(trainHam, axis=0) countsSpam = np.sum(trainSpam, axis=0) p_11=countsSpam/numSpam p_10=countsHam/numHam print(p_11) print(p_10)

predictLables=testFeat[:,vocabInds] matchedPrediction= predictLables == testLabels accuracy= np.sum(matchedPrediction)/len(testLabels) print(accuracy)

vocabInds=[179, 859, 2211] p_wordsSpam= p_11_[vocabInds] p_wordsNotspam=p_10_[vocabInds] p_notwordsSpam=p_01_[vocabInds] p_notwordsNotSpam=p_00_[vocabInds] emailsWord=testFeat[:,vocabInds] n=len(testFeat) predictLables=[] for i in range(n): p_xy1=1 p_xy0=1 for k in range (3): if emailsWord[i,k] == True: p_xy1 = p_xy1 * p_wordsSpam[k] p_xy0 = p_xy0 * p_wordsNotspam[k] else: p_xy1 = p_xy1 * p_notwordsSpam[k] p_xy0 = p_xy0 * p_notwordsNotSpam[k] if p_xy1 > p_xy0: predictLables.append(1) else: predictLables.append(0)

vocabInds= np.arange(W) p_wordsSpam= p_11_[vocabInds] p_wordsNotspam=p_10_[vocabInds] p_notwordsSpam=p_01_[vocabInds] p_notwordsNotSpam=p_00_[vocabInds] emailsWord=testFeat[:,vocabInds] n=len(testFeat) predictLables=[] for i in range(n): p_xy1=0 p_xy0=0 for k in range (W): if emailsWord[i,k] == True: p_xy1 = p_xy1 + math.log(p_wordsSpam[k]) p_xy0 = p_xy0 + math.log(p_wordsNotspam[k]) else: p_xy1 = p_xy1 + math.log(p_notwordsSpam[k]) p_xy0 = p_xy0 + math.log(p_notwordsNotSpam[k]) if p_xy1 > p_xy0: predictLables.append(1) else: predictLables.append(0)