# Kolekce statistik

stats = {}

keys = {}

# Konstanty

# CSV index topics

topic_index = 1

# Celkova spravnost

good = 0

bad = 0

def __init__(self, inputFile, outputFile, trainingCount):

self.create_stats()

self.inputFile = inputFile

self.outputFile = outputFile

self.trainingCount = trainingCount

self.create_topic_map()

self.calculate_statistics()

pass

def create_stats(self):

self.stats['po'] = Statistic('po')

self.stats['pr'] = Statistic('pr')

self.stats['ze'] = Statistic('ze')

self.stats['sp'] = Statistic('sp')

self.stats['ku'] = Statistic('ku')

self.stats['kr'] = Statistic('kr')

self.stats['pc'] = Statistic('pc')

self.stats['ji'] = Statistic('ji')

def create_topic_map(self):

self.keys['po'] = "POLITICS"

self.keys['pr'] = "INDUSTRY"

self.keys['ze'] = "AGRICULTURE"

self.keys['sp'] = "SPORT"

self.keys['ku'] = "CULTURE"

self.keys['kr'] = "CRIME"

self.keys['pc'] = "WEATHER"

self.keys['ji'] = "OTHER"

def get_stats(self, zkratka):

if zkratka not in self.stats.keys():

return None

else:

return self.stats[zkratka]

def calculate_statistics(self):

csv_input = open(self.inputFile, mode="r", encoding='utf-8')

input_reader = csv.reader(csv_input, quotechar=None, delimiter=";")

csv_output = open(self.outputFile, mode="r", encoding='utf-8')

output_reader = csv.reader(csv_output, quotechar=None, delimiter=";")

# Read non training rows from input CSV

input_lines = []

for input_row in input_reader:

if input_reader.line_num > self.trainingCount:

input_lines.append(input_row)

# Read all lines from ouput CSV

output_lines = []

for output_row in output_reader:

output_lines.append(output_row)

# For every line loaded we need to evaluate

for i in range(0, len(output_lines)):

expected = input_lines[i]

real = output_lines[i]

expected_topic = expected[self.topic_index]

real_topic = real[self.topic_index]

if expected_topic == real_topic:

self.good += 1

# Add True NEGATIVE for every other topic

for topicKey in self.stats:

if topicKey != expected_topic:

self.get_stats(topicKey).tn += 1

# Add true positive for correct detection

self.get_stats(expected_topic).tp += 1

else:

self.bad += 1

# Add TRUE

self.get_stats(expected_topic).fp += 1

self.get_stats(real_topic).fn += 1

# Close files after done reading from them

csv_input.close()

csv_output.close()

def print_complete_stat(self):

total = self.good + self.bad

all = [0] * 5

ok = [0] * 5

for i in range(0, len(self.stats.keys())):

stat = self.get_stats(list(self.stats.keys())[i])

topicTweets = stat.fn + stat.tp

ratio = topicTweets / total

precision = stat.precision()

if not math.isnan(precision):

ok[0] += 1.0 * ratio

all[0] += precision * ratio

recall = stat.recall()

if not math.isnan(recall):

ok[1] += 1.0 * ratio

all[1] += recall * ratio

fmeasure = stat.fmeasure()

if not math.isnan(fmeasure):

ok[2] += 1.0 * ratio

all[2] += fmeasure * ratio

error = stat.error()

if not math.isnan(error):

ok[3] += 1.0 * ratio

all[3] += error * ratio

acc = stat.accuracy()

if not math.isnan(acc):

ok[4] += 1.0 * ratio

all[4] += acc * ratio

# After cycle

for y in range(0, len(all)):

all[y] /= ok[y]

print("%s: presnost: %.2f%%, uplnost: %.2f%%, fmira: %.2f%%, chybovost %.2f%%, accuracy: %.2f%%" % (

"CELKOVE", all[0] * 100, all[1] * 100, all[2] * 100, all[3] * 100, all[4]*100))

print("Správně/Celkově: %d/%d" % (self.good, self.good + self.bad))

def print_topic_stats(self):

# Pro vsechny kategorie

for topicKey in self.stats:

# ziskame statistiku kategorie

cat_stat = self.stats[topicKey]

precision = cat_stat.precision() * 100

recall = cat_stat.recall() * 100

fmeasure = cat_stat.fmeasure() * 100

error = cat_stat.error() * 100

accuracy = cat_stat.accuracy() * 100

print("%s: presnost: %.2f%%, uplnost: %.2f%%, fmira: %.2f%%, chybovost %.2f%%, accuracy: %.2f%%" % (

self.translate_key(topicKey), precision, recall, fmeasure, error, accuracy))

def translate_key(self, topicKey):

if topicKey not in self.keys.keys():

return "UNKNOWN"

else: