def random_forest(X_train, y_train, X_test, y_test):
logfile = open("report/logs/random_forest.tex", "w")
print("Random Forest\n")
# Random forest with 100 trees
rf = train_model(X_train,
y_train,
X_test,
y_test,
RandomForestClassifier,
max_depth=1,
n_estimators=100,
random_state=0,
n_jobs=-1)
max_depth = 1
for i in range(2, 10):
tmp = train_model(X_train,
y_train,
X_test,
y_test,
RandomForestClassifier,
max_depth=i,
n_estimators=100,
random_state=0,
n_jobs=-1)
if tmp.score(X_test, y_test) > rf.score(X_test, y_test):
rf = tmp
max_depth = i
# elif tmp.score(X_test, y_test) != rf.score(X_test, y_test):
# break
log(logfile, "Max depth = " + str(max_depth) + "\\\\")
train_model(X_train,
y_train,
X_test,
y_test,
RandomForestClassifier,
logfile,
max_depth=max_depth,
n_estimators=100,
random_state=0,
n_jobs=-1)
plot_learning_curve(rf, X_train, y_train)
logfile.close()
return rf
def naive_bayes(X_train, y_train, X_test, y_test):
logfile = open("report/logs/naive_bayes.tex", "w")
print("Naive Bayes\n")
print("\nGausian")
gnb = train_model(X_train, y_train, X_test, y_test, GaussianNB)
print("\nBernoulli")
bnb = train_model(X_train, y_train, X_test, y_test, BernoulliNB, logfile)
if gnb.score(X_test, y_test) > bnb.score(X_test, y_test):
nb = gnb
else:
nb = bnb
plot_learning_curve(nb, X_train, y_train)
logfile.close()
return nb
def decision_tree(X_train, y_train, X_test, y_test):
logfile = open("report/logs/decision_tree.tex", "w")
print("Decision Tree\n")
print("Max depth = 1")
dt = train_model(X_train,
y_train,
X_test,
y_test,
DecisionTreeClassifier,
max_depth=1,
random_state=0)
max_depth = 1
for i in range(2, 10):
tmp = train_model(X_train,
y_train,
X_test,
y_test,
DecisionTreeClassifier,
max_depth=i,
random_state=0)
if tmp.score(X_test, y_test) > dt.score(X_test, y_test):
dt = tmp
max_depth = i
log(logfile, "Max depth = " + str(max_depth) + "\\\\")
train_model(X_train,
y_train,
X_test,
y_test,
DecisionTreeClassifier,
logfile,
max_depth=max_depth,
random_state=0)
plot_learning_curve(dt, X_train, y_train)
logfile.close()
return dt
def k_nearest_neighbors(X_train, y_train, X_test, y_test):
logfile = open("report/logs/k_nearest_neighbors.tex", "w")
print("K Nearest Neighbors\n")
knn = train_model(X_train,
y_train,
X_test,
y_test,
KNeighborsClassifier,
n_neighbors=1)
n_neighbors = 1
for i in range(2, 10):
tmp = train_model(X_train,
y_train,
X_test,
y_test,
KNeighborsClassifier,
n_neighbors=i)
if tmp.score(X_test, y_test) > knn.score(X_test, y_test):
knn = tmp
n_neighbors = i
log(logfile, "N neighbors = " + str(n_neighbors) + "\\\\")
train_model(X_train,
y_train,
X_test,
y_test,
KNeighborsClassifier,
logfile,
n_neighbors=n_neighbors)
plot_learning_curve(knn, X_train, y_train)
logfile.close()
return knn
def logistic_regression(X_train, y_train, X_test, y_test):
logfile = open("report/logs/log_reg.tex", "w")
print("Logistic Regression\n")
log_reg = train_model(X_train,
y_train,
X_test,
y_test,
LogisticRegression,
logfile,
random_state=0,
solver='liblinear')
plot_learning_curve(log_reg, X_train, y_train)
logfile.close()
return log_reg