def plot_auc_barchart(classifiers, tests):
attitudes = ["proactivo", "reactivo", "agresivo", "provoto"]
results = {}
dfs = {}
for classifier in classifiers:
classifier_name = type(classifier).__name__
for feature_names in tests:
feature = "+".join(
[feature_name.upper()
for feature_name in feature_names] + [classifier_name])
dfs[feature] = dat.getFeaturesDataFrame(*feature_names)
for i, attitude in enumerate(attitudes):
results[attitude] = {}
for classifier in classifiers:
classifier_name = type(classifier).__name__
for feature_names in tests:
feature = "+".join(
[feature_name.upper()
for feature_name in feature_names] + [classifier_name])
df = dfs[feature]
sorted_train_labels = sorted(list(set(df.loc[:, attitude])))
for label in sorted_train_labels:
t = df[(df[attitude] == label)]
if t.shape[0] < 10:
df = df.append([t] * 10)
print("========================={}===========================".
format(attitude.upper()))
data = df.loc[:, [
c for c in df.columns for feature_name in feature_names
if "_" + feature_name.lower() in c
]]
target = df.loc[:, attitude]
results[attitude][feature] = tc.compute_auc_multiclass(
data, target, classifier)
attitudes = ["provoto", "agresivo", "reactivo", "proactivo"]
dist = []
for attitude in attitudes:
dist.append(
go.Bar(
x=list(results[attitude].keys()),
y=[results[attitude][r] for r in results[attitude].keys()],
name=attitude))
fig = go.Figure(data=dist,
layout=go.Layout(title="Average ROC AUC per attitude"))
plotly.offline.plot(
fig,
filename='images/{}-average-roc-auc-per-attitude.html'.format(
classifier_name))
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import train_test_split
import os
import __init__
import matplotlib.pyplot as plt
import numpy as np
import load_dataframe as dat
import config as conf
feature_names = ["big", "w2v", "bow"]
df = dat.getFeaturesDataFrame(*feature_names)
classifier = RandomForestClassifier(n_estimators=100,
n_jobs=-1,
random_state=conf.seed)
attitudes = ["proactivo", "reactivo", "agresivo", "provoto"]
for i, attitude in enumerate(attitudes):
data = df.loc[:, [
c for c in df.columns for feature_name in feature_names
if "_" + feature_name.lower() in c
]]
target = df.loc[:, attitude]
train_data, test_data, train_labels, test_labels = train_test_split(
data, target, test_size=.5, random_state=0)
model = classifier.fit(train_data, train_labels)
importances = model.feature_importances_
std = np.std([tree.feature_importances_ for tree in model.estimators_],
axis=0)
indices = np.argsort(importances)[::-1]
max_features = 25
plt.figure()
title = '{}+RandomForest feature importance for attitude {}'.format(
"+".join([feature_name.upper() for feature_name in feature_names]),
"Attitude '{}'".format(attitude),
save=conf.images_dir,
plot=axarr[i])
plt.setp([a.get_yticklabels() for a in axarr[1:]], visible=False)
f.subplots_adjust(hspace=0.3)
plt.suptitle('{}+{} for all attributes'.format(
"+".join([feature_name.upper() for feature_name in feature_names]),
classifier_name),
y=1)
f.text(0.5, 0.01, 'False Positive Rate', ha='center')
f.text(0.09, 0.5, 'True Positive Rate', va='center', rotation='vertical')
plt.savefig(os.path.join(
conf.images_dir, '{}+{} for all attributes.png'.format(
"+".join([feature_name.upper() for feature_name in feature_names]),
classifier_name)),
bbox_inches='tight')
# plt.show()
if (__name__ == "__main__"):
classifiers = [
BernoulliNB()
# ,RandomForestClassifier(n_estimators=300,n_jobs=-1, random_state=conf.seed)
]
tests = [["bow"], ["big"], ["w2v"], ["bow", "big"], ["bow", "w2v"],
["big", "w2v"], ["big", "w2v", "bow"]]
for classifier in classifiers:
for test in tests:
AUC(classifier, data.getFeaturesDataFrame(*test), test)
# AUC(BernoulliNB(), bow.ExtractW2V)
# AUC(BernoulliNB(), bow.ExtractBOW, w2v.ExtractW2V, big.ExtractBIG)
def plot_confusion_matrix(classifiers, tests):
for classifier in classifiers:
for test in tests:
confusion_matrix(classifier, dat.getFeaturesDataFrame(*test), test)
# Plot ROC curves for the multiclass problem # Compute macro-average ROC curve and ROC area # First aggregate all false positive rates all_fpr = np.unique(np.concatenate([fpr[i] for i in range(n_classes)])) # Then interpolate all ROC curves at this points mean_tpr = np.zeros_like(all_fpr) for i in range(n_classes): mean_tpr += np.interp(all_fpr, fpr[i], tpr[i]) # Finally average it and compute AUC mean_tpr /= n_classes fpr["macro"] = all_fpr tpr["macro"] = mean_tpr roc_auc["macro"] = auc(fpr["macro"], tpr["macro"]) return roc_auc["macro"] if(__name__=="__main__"): classifiers=[ BernoulliNB(), RandomForestClassifier(n_estimators=300,n_jobs=-1, random_state=conf.seed) ] tests=[["bow"],["big"],["w2v"],["bow","big"],["bow","w2v"],["big","w2v"],["big","w2v","bow"]] for classifier in classifiers: for test in tests: get_accuracy(classifier, dat.getFeaturesDataFrame(*test), test)