[SOLVED] import numpy as np

import numpy as np
import matplotlib.pyplot as plt
from getDataset import getDataSet
from sklearn.linear_model import LogisticRegression

Copyright By Assignmentchef assignmentchef

# Starting codes

# Fill in the codes between %PLACEHOLDER#start and PLACEHOLDER#end

# step 1: generate dataset that includes both positive and negative samples,
# where each sample is described with two features.
# 250 samples in total.

[X, y] = getDataSet()# note that y contains only 1s and 0s,

# create figure for all charts to be placed on so can be viewed together
fig = plt.figure()

def func_DisplayData(dataSamplesX, dataSamplesY, chartNum, titleMessage):
idx1 = (dataSamplesY == 0).nonzero()# object indices for the 1st class
idx2 = (dataSamplesY == 1).nonzero()
ax = fig.add_subplot(1, 3, chartNum)
# no more variables are needed
plt.plot(dataSamplesX[idx1, 0], dataSamplesX[idx1, 1], r*)
plt.plot(dataSamplesX[idx2, 0], dataSamplesX[idx2, 1], b*)
# axis tight
ax.set_xlabel(x_1)
ax.set_ylabel(x_2)
ax.set_title(titleMessage)

# plotting all samples
func_DisplayData(X, y, 1, All samples)

# number of training samples
nTrain = 120

######################PLACEHOLDER 1#start#########################
# write you own code to randomly pick up nTrain number of samples for training and use the rest for testing.
# WARNIN: do not use the scikit-learn or other third-party modules for this step

#maxIndex = len(X)
#randomTrainingSamples = np.random.choice(maxIndex, nTrain, replace=False)
trainX = #training samples
trainY =# labels of training samplesnTrain X 1

testX = # testing samples
testY = # labels of testing samples nTest X 1

####################PLACEHOLDER 1#end#########################

# plot the samples you have pickup for training, check to confirm that both negative
# and positive samples are included.
func_DisplayData(trainX, trainY, 2, training samples)
func_DisplayData(testX, testY, 3, testing samples)

# show all charts
plt.show()

#step 2: train logistic regression models

######################PLACEHOLDER2 #start#########################
# in this placefolder you will need to train a logistic model using the training data: trainX, and trainY.
# please delete these coding lines and use the sample codes provided in the folder codeLogit
logReg = LogisticRegression(fit_intercept=True, C=1e15) # create a model
logReg.fit(trainX, trainY)# training
coeffs = logReg.coef_ # coefficients
intercept = logReg.intercept_ # bias
bHat = np.hstack((np.array([intercept]), coeffs))# model parameters
######################PLACEHOLDER2 #end #########################

# step 3: Use the model to get class labels of testing samples.

######################PLACEHOLDER3 #start#########################
# codes for making prediction,
# with the learned model, apply the logistic model over testing samples
# hatProb is the probability of belonging to the class 1.
# y = 1/(1+exp(-Xb))
# yHat = 1./(1+exp( -[ones( size(X,1),1 ), X] * bHat )); ));
# WARNING: please DELETE THE FOLLOWING CODEING LINES and write your own codes for making predictions
xHat = np.concatenate((np.ones((testX.shape[0], 1)), testX), axis=1)# add column of 1s to left most->130 X 3
negXHat = np.negative(xHat)# -1 multiplied by matrix -> still 130 X 3
hatProb = 1.0 / (1.0 + np.exp(negXHat * bHat))# variant of classification -> 130 X 3
# predict the class labels with a threshold
yHat = (hatProb >= 0.5).astype(int)# convert bool (True/False) to int (1/0)
#PLACEHOLDER#end

######################PLACEHOLDER 3 #end #########################

# step 4: evaluation
# compare predictions yHat and and true labels testy to calculate average error and standard deviation
testYDiff = np.abs(yHat testY)
avgErr = np.mean(testYDiff)
stdErr = np.std(testYDiff)

print(average error: {} ({}).format(avgErr, stdErr))

######################PLACEHOLDER4 #start#########################
#Please add codes to calculate the ROC curves and AUC of the trained model over all the testing samples

######################PLACEHOLDER4 #end #########################

CS: assignmentchef QQ: 1823890830 Email: [email protected]