def runAll(path, predictMethodList, predictMethod):
users = orig.loadAndUpdateFeatures(path)
featureList = orig.featureList()
featureList.addByRegex(["action_", "num_of_devices", "total_time"], users)
category = 'country_destination'
X_byteDF, y = orig.getXbyte(users, featureList.get(), category)
return runClf2net(predictMethodList, predictMethod, X_byteDF, y)
def __init__(self, args, genes=None, mutationFactor=0.01):
self.args = args
featureListClass = orig.featureList()
featureListClass.addByRegex(['action_','num_of_devices'],self.args['users'])
self.featureList = featureListClass.get()
self.category = 'country_destination'
self.mutationFactor = mutationFactor
self.fitScoreNorm = 0
self.fitScore = 0
self.fit = 0
if genes is None :
self.genes = []
for i in range(len(self.featureList)):
if getRandTrueFalse():
self.genes.append(self.featureList[i])
else :
self.genes = genes
# self.predictMethod = LogisticRegression()
self.predictMethod = tree.DecisionTreeClassifier()
def __init__(self, args, genes=None, mutationFactor=0.01):
self.args = args
featureListClass = orig.featureList(
usersCol=self.args['users'].columns)
self.featureList = featureListClass.get()
self.category = 'country_destination'
self.mutationFactor = mutationFactor
self.fitScoreNorm = 0
self.fitScore = 0
self.fit = 0
if genes is None:
self.genes = []
for i in range(len(self.featureList)):
if getRandTrueFalse():
self.genes.append(self.featureList[i])
else:
self.genes = genes
# self.predictMethod = LogisticRegression()
self.predictMethod = self.args['predictMethod']
def __init__(self, args, genes=None, mutationFactor=0.01):
self.args = args
featureListClass = orig.featureList(
usersCol=self.args['users'].columns)
self.featureList = featureListClass.get()
self.category = 'country_destination'
self.mutationFactor = mutationFactor
self.fitScoreNorm = 0
self.fitScore = 0
self.fit = 0
self.maxNumOfNurons = 100
self.minAlpaP2 = -6
self.maxAlpaP2 = 4
if genes is None:
self.genes = self.getRandomGenes()
else:
self.genes = genes
self.predictMethod = MLPClassifier(solver='lbfgs',
alpha=self.genes[0],
hidden_layer_sizes=tuple(
self.genes[1:]))
# -*- coding: utf-8 -*-
"""
Created on Tue Mar 7 17:37:46 2017
@author: Alon
"""
import basicLib.loadAndTest as orig
from sklearn.neural_network import MLPClassifier
import clfNdfOtherAll as clfAll
from sklearn import tree
users = orig.loadAndUpdateFeatures(
'../input/users_2014_actions_combined_device.csv')
featureListClass = orig.featureList()
featureListClass.addByRegex(['action_', 'num_of_devices'], users)
featureList = featureListClass.get()
category = 'country_destination'
predictMethod1 = MLPClassifier(solver='lbfgs', alpha=1e-5)
predictMethod2 = MLPClassifier(solver='lbfgs', alpha=1e-5)
predictMethod3 = tree.DecisionTreeClassifier()
clfNdfOtherAllClass = clfAll.clfNdfOtherAll(predictMethod1, predictMethod3)
prediction, fit, y_test = orig.fitPredictAndTest(users,
featureList,
category,
clfNdfOtherAllClass,
random_state=1)
print(fit)
print(clfNdfOtherAllClass.accuracy_score(y_test))
# -*- coding: utf-8 -*-
"""
Created on Thu Mar 9 11:33:24 2017
@author: Alon
"""
import featureRapperClf as frc
import basicLib.loadAndTest as orig
from sklearn import tree
import time
predictMethod = tree.DecisionTreeClassifier()
users = orig.loadAndUpdateFeatures('../input/users_2014_sessions_norm.csv')
featureListClass = orig.featureList(usersCol=users.columns)
featureListAll = featureListClass.get()
category = 'country_destination'
featureList = [
'action_set_password_##_submit_##_set_password',
'action_authenticate_##_submit_##_login'
]
featureRapper = frc.featureRapperClf(predictMethod, featureList)
startRun = time.clock()
prediction, fit = orig.fitPredictAndTest(users,
featureListAll,
category,
featureRapper,
random_state=1)