def formatFeatures():
data = []
# Isolating feature House
y = []
for _ in range(lenFeatures):
data.append([])
for row in rawdata:
if ml.isFormatted(row):
if row[1] == 'Gryffindor':
y.append(1.0)
elif row[1] == 'Ravenclaw':
y.append(2.0)
elif row[1] == 'Slytherin':
y.append(3.0)
elif row[1] == 'Hufflepuff':
y.append(4.0)
for i in range(lenFeatures):
data[i].append(float(row[i + indexFeatures]))
# Normalize
for i in range(lenFeatures):
minV, maxV = ml.getMinMax(data[i])
data[i] = ml.normalizeData(data[i], minV, maxV)
return data, y
def formatFeatures():
data = []
for _ in range(lenFeatures):
data.append([])
for row in rawdata:
for i in range(lenFeatures):
if row[i + indexFeatures] == '':
data[i].append(0.0)
else:
data[i].append(float(row[i + indexFeatures]))
# Normalize
for i in range(lenFeatures):
minV, maxV = ml.getMinMax(data[i])
data[i] = ml.normalizeData(data[i], minV, maxV)
return data
def predict():
f = open('assets/houses.csv', 'w')
f.write('Index,Hogwarts House\n')
distribution = [[], [], [], []]
for i, studMarks in enumerate(X):
probs = []
for _, weights in enumerate(housesWeights):
z = np.dot(studMarks, weights)
probs.append(sigmoid(z))
_, maxV = ml.getMinMax(probs)
house = ''
for j in range(len(probs)):
if probs[j] == maxV:
house = houses[j]
if house == 'Gryffindor':
distribution[0].append('Gryffindor')
elif house == 'Ravenclaw':
distribution[1].append('Ravenclaw')
elif house == 'Slytherin':
distribution[2].append('Slytherin')
elif house == 'Hufflepuff':
distribution[3].append('Hufflepuff')
f.write(str(i) + ',' + house + '\n')
return distribution
marks = []
# Get data
for i in range(lenFeatures + 1):
marks.append([])
for row in rawdata:
if ml.isFormatted(row):
marks[lenFeatures].append(row[1])
for i in range(lenFeatures):
marks[i].append(float(row[i + indexFeatures]))
# Normalize
for i in range(lenFeatures):
minV, maxV = ml.getMinMax(marks[i])
marks[i] = ml.normalizeData(marks[i], minV, maxV)
# Prepare data for pair plot
pairplot = []
for i in range(len(marks[0])):
tmp = []
for j in range(lenFeatures + 1):
tmp.append(marks[j][i])
pairplot.append(tmp)
# Plot
pairplot = pd.DataFrame(pairplot, columns=features)
sns_plot = sns.pairplot(pairplot, size=2.5, hue=features[lenFeatures])
sns_plot.savefig("pair_plot.png")
for row in rawdata:
if ml.isFormatted(row):
tmp = gryf
if row[1] == 'Ravenclaw':
tmp = raven
elif row[1] == 'Slytherin':
tmp = slyth
elif row[1] == 'Hufflepuff':
tmp = huffle
for j in range(lenFeatures):
tmp[j].append(row[j + indexFeatures])
# Get all stats
for i, house in enumerate(houses):
for j, row in enumerate(house):
house[j] = ml.formatData(row)
minV, maxV = ml.getMinMax(house[j])
house[j] = ml.normalizeData(house[j], minV, maxV)
mean = ml.getMean(house[j])
housesStd[i][j] = ml.getStd(house[j], mean)
stdMean[j] += housesStd[i][j]
for i, mean in enumerate(stdMean):
stdMean[i] /= lenHouses
# Get Std
for i, house in enumerate(housesStd):
for j, row in enumerate(house):
featuresStd[j] += (row - stdMean[j]) * (row - stdMean[j])
for i, std in enumerate(featuresStd):
featuresStd[i] /= lenHouses - 1