def result(train_fraction, test_fraction):
all_data = main.get_data(train_fraction,
test_fraction) # 0 = train 1 = test
testing_data = all_data[1]
all_models = train_bayes.get_models(all_data[0])
tp = 0
fp = 0
tn = 0
fn = 0
#p = high
#n = low
for i in testing_data:
if train_bayes.predict(all_models, i[0:len(i) - 1]) == i[-1]:
if i[-1] == 3:
tp += 1
else:
tn += 1
else:
if i[-1] == 3:
fn += 1
else:
fp += 1
# tpr = high_correct/(high_correct+high_wrong)
# far =
confusion_matrix = [[tn, fp], [fn, tp]]
return confusion_matrix
def plotRoc():
tprList = []
farList = []
ac_list = []
ml = []
x = []
runs = 100
aux_list = (main.get_shuffle_data()) # 0 = train 1 = test
part = main.partition_data(aux_list, 0, int(len(aux_list) * 0.3))
all_models = train_bayes.get_models(part[0])
for i in range(runs):
x.append(i)
cm = roc_result(all_models, part[1], 1 / (runs - i))
tp = cm[1][1]
fp = cm[0][1]
tn = cm[0][0]
fn = cm[1][0]
ac_list.append(((cm[0][0] + cm[1][1]) /
(cm[0][0] + cm[1][1] + cm[0][1] + cm[1][0])) * 100)
# tprList.append(tp/(tp+fn))
# farList.append(fp/(fp+tn))
ml.append((fp / (fp + tn), tp / (tp + fn)))
for i in range(runs):
farList.append(ml[i][0])
tprList.append(ml[i][1])
plt.figure(2)
plt.plot(farList, tprList)
plt.ylabel('TPR')
plt.title("100 Run FAR vs TPR")
plt.xlabel('FAR')
plt.show()
def kFold():
accuracy_list = []
aux_list = (main.get_shuffle_data()) # 0 = train 1 = test
for i in range(0, len(aux_list), int(len(aux_list) / 5)):
if i + int(len(aux_list) / 5) <= len(aux_list):
part = main.partition_data(aux_list, i, i + int(len(aux_list) / 5))
all_models = train_bayes.get_models(part[0])
cm = result1(all_models, part[1])
accuracy_list.append(
((cm[0][0] + cm[1][1]) /
(cm[0][0] + cm[1][1] + cm[0][1] + cm[1][0])) * 100)
normal_acc = calc_accuracy(result(0.7, 0.3))
return [accuracy_list, normal_acc]
def compare_thirty_fifty():
thirty = []
fifty = []
t = []
for i in range(30):
aux_list = (main.get_shuffle_data()) # 0 = train 1 = test
part = main.partition_data(aux_list, 0, int(len(aux_list) * 0.3))
all_models = train_bayes.get_models(part[0])
cm1 = result1(all_models, part[1])
part = main.partition_data(aux_list, 0, int(len(aux_list) * 0.5))
all_models = train_bayes.get_models(part[0])
cm2 = result1(all_models, part[1])
thirty.append(calc_accuracy(cm1))
fifty.append(calc_accuracy(cm2))
t.append(i)
plt.figure(3)
plt.plot(t, thirty, label='70-30 split')
plt.plot(t, fifty, label='50-50 split')
plt.ylabel('Accuracy')
plt.title("30 Run accuracy 70-30 and 50-50 split")
plt.xlabel('Run number')
plt.legend(bbox_to_anchor=(1, 1), bbox_transform=plt.gcf().transFigure)
plt.show()
if __name__ == "__main__":
s = """
low(predicted) high(predicted)
low(actual) tn fp
high(actual) fn tp
"""
print(s)
print("Part A. Simple Test 70% training results:")
aux_list = (main.get_shuffle_data()) # 0 = train 1 = test
part = main.partition_data(aux_list, 0, int(len(aux_list) * 0.3))
all_models = train_bayes.get_models(part[0])
train_bayes.save_model(all_models)
cm1 = result1(all_models, part[1])
print_ac(cm1)
print("Accuracy graph for 100 runs")
plot_hundred_runs()
compare_thirty_fifty()
kfold_list = kFold()
print("Mean 5 fold = ", str(train_bayes.listMean(kfold_list[0])))
print("Std dev 5 fold = ", str(train_bayes.sigma(kfold_list[0])))
print("70-30 accuracy = ", str(kfold_list[1]))
# print(kFold())
# result(0.7,0.3)
# result(0.5,0.5)