def test(test_data, test_labels, type_of_data, algorithm):
if type_of_data == 1:
weights_perceptron = np.load('digits_perceptron_weights.npy')
features_train = np.load('digits_knn_features.npy')
height = 28
width = 28
classes = 10
else:
weights_perceptron = np.load('faces_perceptron_weights.npy')
features_train = np.load('faces_knn_features.npy')
height = 70
width = 60
classes = 2
samples, sample_lines = readfile(test_data, type_of_data)
samples = getsamples(samples, sample_lines, height, width)
labels, label_lines = readfile(test_labels, type_of_data)
labels = getlabels(labels)
if type_of_data == 1:
feature_matrix = get_features_for_digits(samples)
if algorithm == 'perceptron':
test_multiclass_perceptron(feature_matrix, labels, classes, weights_perceptron)
if algorithm == 'knn':
test_knn(feature_matrix, labels, features_train, 1)
if algorithm == 'naivebayes':
bayes(feature_matrix, labels, features_train, 1)
else:
feature_matrix = get_features_for_faces(samples)
if algorithm == 'perceptron':
test_binary_perceptron(feature_matrix, labels, classes, weights_perceptron)
if algorithm == 'knn':
test_knn(feature_matrix, labels, features_train, 2)
if algorithm == 'naivebayes':
bayes(feature_matrix, labels, features_train, 2)
def bayes_naive(self, predictData, trainData):
h = hp()
nb = bayes()
accuracy = []
precision = []
recall = []
f_score = []
for i in range(len(trainData)):
tmp = None
predictData = trainData[i]
tmp = [lt for j, lt in enumerate(trainData) if j != i]
td = h.convertToList(tmp)
classPriorProbabilities = nb.findClassPriorProbability(td)
classes = nb.segregateClasses(td)
occurences, means, stdDev = nb.findDescriptorPosteriorProbabilites(
classes, td)
nb.classify(predictData, classPriorProbabilities, occurences,
means, stdDev)
truePositives, trueNegatives, falsePositives, falseNegatives = h.findParams(
predictData)
accuracy.append(
h.findAccuracy(truePositives, trueNegatives, falsePositives,
falseNegatives))
tmpPrecision = h.findPrecision(truePositives, trueNegatives,
falsePositives, falseNegatives)
tmpRecall = h.findRecall(truePositives, trueNegatives,
falsePositives, falseNegatives)
precision.append(tmpPrecision)
recall.append(tmpRecall)
f_score.append(h.findFMeasure(tmpPrecision, tmpRecall))
return accuracy, precision, recall, f_score
def bayes_naive(self, predictData, trainData):
h = hp()
nb = bayes()
matrix = defaultdict(list)
pd = [pt for pt in predictData]
# for i in range(len(trainData)):
tmp = [lt for j, lt in enumerate(trainData)]
td = h.convertToList(tmp)
classPriorProbabilities = nb.findClassPriorProbability(td)
classes = nb.segregateClasses(td)
occurences, means, stdDev = nb.findDescriptorPosteriorProbabilites(classes, td)
nb.classify(predictData, classPriorProbabilities, occurences, means, stdDev)
return predictData
def getFitness(population):
fitness = []
print population
for i in range(len(population)):
current_no_of_features = 0
for j in range(no_of_features):
if population[i][j] == 1:
current_no_of_features += 1
new_data_frame = dropColumns(population[i], data_frame)
accuracy = bayes(new_data_frame)
#print "Accuracy",accuracy
#Giving more importance to states having less no of features
fitness.append(0.99 * accuracy + 0.01 *
(no_of_features - current_no_of_features))
return fitness
def bayes_naive_demo(self, predictData, trainData):
h = hp()
nb = bayes()
classPriorProbabilities = nb.findClassPriorProbability(trainData)
classes = nb.segregateClasses(trainData)
occurences, means, stdDev = nb.findDescriptorPosteriorProbabilites(
classes, trainData)
probabilities = nb.classify_demo(predictData, classPriorProbabilities,
occurences, means, stdDev)
maxProb = float('-inf')
classKey = -1
for key in probabilities:
print("P(X|H{})*P(H{}) = {}".format(key, key, probabilities[key]))
if probabilities[key] > maxProb:
maxProb = probabilities[key]
classKey = key
print("This test data record belongs to: Class {}".format(classKey))
import pandas as pd
import random as random
import math as math
import numpy as np
from naive_bayes import bayes
file_name = raw_input("Enter data file location: ")
#col_names = ['pregnant', 'glucose', 'bp', 'skin', 'insulin', 'bmi', 'pedigree', 'age', 'label'] #features of the pima data set
data_frame = pd.read_csv(file_name, header=None)
no_of_features = len(data_frame.columns) - 1
print no_of_features, " columns"
original_accuracy = bayes(data_frame) #get accuracy with all features present
print "Original accuracy", original_accuracy
uniformProbability = 0.5
mutationProbability = 0.1
#Create a random individual
def createIndividual():
all_zero = True
while True:
genes = np.random.choice([0, 1],
size=(no_of_features, ),
p=[4. / 5, 1. / 5])
for i in range(len(genes)):
if genes[i] == 1:
all_zero = False
break
if all_zero == False:
break