General features selection based on certain machine learning algorithm and evaluation methods
Divesity, Flexible and Easy to use
More features selection method will be included in the future!
pip3 install MLFeatureSelectionModulus for selecting features based on greedy algorithm (from MLFeatureSelection import sequence_selection)
Modulus for removing features based on features importance (from MLFeatureSelection import importance_selection)
Modulus for removing features based on correlation coefficient (from MLFeatureSelection import coherence_selection)
Modulus for reading the features combination from log file (from MLFeatureSelection.tools import readlog)
-- https://github.com/duxuhao/rong360-season2
-- https://github.com/duxuhao/JData-2018
-- https://github.com/duxuhao/IJCAI-2018-2
from MLFeatureSelection import sequence_selection
from sklearn.linear_model import LogisticRegression
sf = sequence_selection.Select(Sequence = True, Random = True, Cross = False)
sf.ImportDF(df,label = 'Label') #import dataframe and label
sf.ImportLossFunction(lossfunction, direction = 'ascend') #import loss function handle and optimize direction, 'ascend' for AUC, ACC, 'descend' for logloss etc.
sf.InitialNonTrainableFeatures(notusable) #those features that is not trainable in the dataframe, user_id, string, etc
sf.InitialFeatures(initialfeatures) #initial initialfeatures as list
sf.GenerateCol() #generate features for selection
sf.SetFeatureEachRound(50, False) # set number of feature each round, and set how the features are selected from all features (True: sample selection, False: select chunk by chunk)
sf.clf = LogisticRegression() #set the selected algorithm, can be any algorithm
sf.SetLogFile('record.log') #log file
sf.run(validate) #run with validation function, validate is the function handle of the validation function, return best features combinationfrom MLFeatureSelection import importance_selection
import xgboost as xgb
sf = importance_selection.Select()
sf.ImportDF(df,label = 'Label') #import dataframe and label
sf.ImportLossFunction(lossfunction, direction = 'ascend') #import loss function and optimize direction
sf.InitialFeatures() #initial features, input
sf.SelectRemoveMode(batch = 2)
sf.clf = xgb.XGBClassifier()
sf.SetLogFile('record.log') #log file
sf.run(validate) #run with validation function, return best features combinationfrom MLFeatureSelection import coherence_selection
import xgboost as xgb
sf = coherence_selection.Select()
sf.ImportDF(df,label = 'Label') #import dataframe and label
sf.ImportLossFunction(lossfunction, direction = 'ascend') #import loss function and optimize direction
sf.InitialFeatures() #initial features, input
sf.SelectRemoveMode(batch = 2)
sf.clf = xgb.XGBClassifier()
sf.SetLogFile('record.log') #log file
sf.run(validate) #run with validation function, return best features combinationfrom MLFeatureSelection.tools import readlog
logfile = 'record.log'
logscore = 0.5 #any score in the logfile
features_combination = readlog(logfile, logscore)from MLFeatureSelection.tools import readlog, filldf
def add(x,y):
return x + y
def substract(x,y):
return x - y
def times(x,y):
return x * y
def divide(x,y):
return x/y
def sq(x,y):
return x ** 2
CrossMethod = {'+':add,
'-':substract,
'*':times,
'/':divide,
} # set your own cross method
df = pd.read_csv('XXX')
logfile = 'record.log'
logscore = 0.5 #any score in the logfile
features_combination = readlog(logfile, logscore)
df = filldf(df, features_combination, CrossMethod)define your own:
validate: validation method in function , ie k-fold, last time section valdate, random sampling validation, etc
lossfunction: model performance evaluation method, ie logloss, auc, accuracy, etc
def validate(X, y, features, clf, lossfunction):
"""define your own validation function with 5 parameters
input as X, y, features, clf, lossfunction
clf is set by SetClassifier()
lossfunction is import earlier
features will be generate automatically
function return score and trained classfier
"""
clf.fit(X[features],y)
y_pred = clf.predict(X[features])
score = lossfuntion(y_pred,y)
return score, clf
def lossfunction(y_pred, y_test):
"""define your own loss function with y_pred and y_test
return score
"""
return np.mean(y_pred == y_test)Multiple processing can be set in validate function when you are doing N-fold.
More examples are added in example folder include:
Demo contain all modulus can be found here (demo)
Simple Titanic with 5-fold validation and evaluated by accuracy (demo)
Demo for S1, S2 score improvement in JData 2018 predict purchase time competition (demo)
Demo for IJCAI 2018 CTR prediction (demo)