def get_model(classifier_name,
training_data,
target_data,
model_settings=None):
"""
Function returns a trained model based on the csv file and features chosen.
Creation and training of models has been separated into their own file:
models.py
Arguments:
training_data: the data the training is performed on
target_data: the targets for supervised learning
model_settings: dict of settings for the classifier
Returns:
model: the trained model
"""
model = ((models.linreg(training_data, target_data)) if
(classifier_name == "linreg") else
(models.svm(training_data, target_data, model_settings)) if
(classifier_name == "svm") else
(models.cart(training_data, target_data)) if
(classifier_name == "cart") else
(models.knn(training_data, target_data, model_settings)) if
(classifier_name == "knn") else
(models.lda(training_data, target_data, model_settings)) if
(classifier_name == "lda") else
(models.nb(training_data, target_data)) if
(classifier_name == "nb") else
(models.lr(training_data, target_data)) if
(classifier_name == "lr") else None)
return model
def main(argv):
param_fpath = sys.argv[1]
# prepare training data
X, y = get_train_data(param_fpath) # X: feature list y: label list
# train
print '\nBegin Running Logistic Regression...'
lr_acc = lr(X, y)
print "LR Accuracy: " + str(lr_acc)
print '\nBegin Running KNN...'
knn_acc = knn(X, y)
print "KNN Accuracy: " + str(knn_acc)
print '\nBegin Running GNB...'
gnb_acc = gnb(X, y)
print "GNB Accuracy: " + str(gnb_acc)
print '\nBegin Running MNB...'
mnb_acc = mnb(X, y)
print "MNB Accuracy: " + str(mnb_acc)
print '\nBegin Running BNB...'
bnb_acc = bnb(X, y)
print "BNB Accuracy: " + str(bnb_acc)
print '\nBegin Running Decision Tree...'
dtree_acc = dtree(X, y)
print "Decision Tree Accuracy: " + str(dtree_acc)
print '\nBegin Running SVM...'
svm_acc = svm(X, y)
print "SVM Accuracy: " + str(svm_acc)
def training_KFold(index, data, name):
_min = sys.maxsize
_model = None
for train, validation in index:
train_set = data[train]
validation_set = data[validation]
training_X, training_y = models.features_labels(train_set)
validation_X, validation_y = models.features_labels(validation_set)
if name == 'gbr':
model, error = models.gbr(training_X, training_y, validation_X,
validation_y)
if name == 'svm':
model, error = models.svm(training_X, training_y, validation_X,
validation_y)
if name == 'rfr':
model, error = models.rfr(training_X, training_y, validation_X,
validation_y)
if name == 'mlr':
model, error = models.mlr(training_X, training_y, validation_X,
validation_y)
if _min > error:
_min = error
_model = model
print("update model")
print(error)
print()
return _model
def runAllModels(X_train,X_test,y_test,y_train):
y_pred= xgb(X_train,y_train,X_test)
plotter(y_test,y_pred)
cm,fscore,a=evaluate(y_test,y_pred)
y_pred = rf(X_train,y_train,X_test)
plotter(y_test,y_pred)
cm,fscore,a=evaluate(y_test,y_pred)
y_pred = nn(X_train,y_train,X_test)
plotter(y_test,y_pred)
cm,fscore,a=evaluate(y_test,y_pred)
y_pred = svm(X_train,y_train,X_test)
plotter(y_test,y_pred)
cm,fscore,a=evaluate(y_test,y_pred)
def runModels(testdata, testlabel):
res = models.knn(testdata, testlabel)
print("knn: ", res.mean())
res = models.svm(testdata, testlabel)
print("svm: ", res.mean())
res = models.decisionTree(testdata, testlabel)
print("decision tree: ", res.mean())
res = models.naiveBayes(testdata, testlabel)
print("Gaussian NB: ", res.mean())
res = models.randomForest(testdata, testlabel)
print("random forest: ", res.mean())
res = models.nearestCentroid(testdata, testlabel)
print("nearest centroid: ", res.mean())
res = models.extraTree(testdata, testlabel)
print("extra tree: ", res.mean())
res = models.extraTrees(testdata, testlabel)
print("extra trees: ", res.mean())
def training(data, name, arg):
training_set = data[:3 * len(data) // 4]
validation_set = data[3 * len(data) // 4:]
training_X, training_y = models.features_labels(training_set)
validation_X, validation_y = models.features_labels(validation_set)
if name == 'gbr':
model, error = models.gbr(training_X,
training_y,
validation_X,
validation_y,
n_estimators=10,
min_samples_split=arg)
if name == 'svm':
model, error = models.svm(training_X, training_y, validation_X,
validation_y)
if name == 'rfr':
model, error = models.rfr(training_X, training_y, validation_X,
validation_y)
if name == 'mlr':
model, error = models.mlr(training_X, training_y, validation_X,
validation_y)
print(error)
print()
return model
training_file, development_file)
print("Word length training: " + str(training_performance))
print("Word length development: " + str(development_performance))
frequency_training_performance, frequency_development_performance = \
word_frequency_threshold(training_file, development_file, counts)
print("Word frequency training: " + str(frequency_training_performance))
print("Word frequency development: " +
str(frequency_development_performance))
nb_development_performance = naive_bayes(training_file, development_file,
counts)
print("Naive bayes development: " + str(nb_development_performance))
lr_development_performance = logistic_regression(training_file,
development_file, counts)
print("Logistic Regression development: " +
str(lr_development_performance))
svm_development_performance = models.svm(training_file, development_file,
test_file, counts)
print("Support vector machine development: " +
str(svm_development_performance))
random_forest_development_performance = models.random_forest(
training_file, development_file, test_file, counts)
print("Random forest development: " +
str(random_forest_development_performance))
# tune_parameter(training_file, development_file, counts)
import pandas as pd
import preprocessingfile as preprocess
import models
data = 'pc2.csv'
original_data, original_X, original_Y, combined_training_data, x_train1, x_train2, x_train, x_test, x_val, y_train1, y_train2, y_train, y_test, y_val = preprocess.my_sdp_preprocessor(
data)
all_data = [
original_data, original_X, original_Y, combined_training_data, x_train1,
x_train2, x_train, x_test, x_val, y_train1, y_train2, y_train, y_test,
y_val
]
cnn_clf = models.cnn(*all_data)
svm_clf = models.svm(*all_data)
rf_clf = models.random_forest(*all_data)
nn_clf = models.NN(*all_data)
from sklearn.metrics import *
def print_accuracy(model): #nn,cnn,svm,clf
if (model == nn_clf):
y_pred_on_val = model.predict(x_val) > 0.5
y_pred_on_test = model.predict(x_test) > 0.5
elif (model == cnn_clf):
x_val_matrix = x_val.values
x_val1 = x_val_matrix.reshape(x_val_matrix.shape[0], 1,
len(x_val.columns), 1)
y_pred_on_val = model.predict(x_val1) > 0.5