def numericCol(table, structureTextFile):
structure = pd.read_csv(structureTextFile,
sep=" ",
names=['type', 'feature', 'data'])
column = []
headers = getColumnTitles(table)
for i in range(structure.shape[0]):
if 'NUMERIC' in structure.loc[i]['data']:
column += [headers[i]]
return column
def allArraysOfFetures(table, classCol):
"""
:param table:
:param classCol:
:return: dict, keys is tuple(feature,'yes'/'no'),values is list of probabilities of the values of the key
"""
thisDict = {}
for i in getColumnTitles(table):
if i not in classCol:
for j in valuesType(table, classCol):
thisDict[i, j] = pArrayByFeature(table, i, j, classCol)
# print(pArrayByFeature(train,i,j,classCol))
return thisDict
def numericCol(table, structureTextFile):
"""
:param table:
:param structureTextFile: path
:return: list of column which is numeric by the structure file
"""
structure = pd.read_csv(structureTextFile, sep=" ", names=['type', 'feature', 'data'])
column = []
headers = getColumnTitles(table)
for i in range(structure.shape[0]):
if 'NUMERIC' in structure.loc[i]['data']:
column += [headers[i]]
return column
def naiveBayes(test, train, structFile):
"""
print the accuracy of the model by test file
:param test:
:param train:
:param structure:
"""
thisDict = allArraysOfFetures(train, 'class')
rows = test.shape[0]
match_yes = 0
match_no = 0
fail_no = 0
fail_yes = 0
# save model to file
filename = 'naiveBayes_model.sav'
joblib.dump(thisDict, filename)
column = getColumnTitles(test)[:-1] # clean 'class' column
for _ in range(rows):
noPar = 1
yesPar = 1
for col in column:
try:
index = valuesType(train, col).index(test.iloc[_][col])
yesPar *= thisDict[(col, 'yes')][index]
noPar *= thisDict[(col, 'no')][index]
except:
continue
if yesPar > noPar:
if test.iloc[_]['class'] == 'yes':
match_yes += 1
else:
fail_yes += 1
else:
if test.iloc[_]['class'] == 'no':
match_no += 1
else:
fail_no += 1
#print('naiveBayes accuracy:', ((match_yes+match_no)) / rows), '%')
Eval(match_yes, match_no, fail_yes, fail_no)