if ver_score[x-1][-1] > score:
score = ver_score[x-1][-1]
max_x = x
max_y = y
print "X: ", x, ". Y: ", y, ":"
print "Training Score: ", train_score[x-1][-1], ". Verification Score: ", \
ver_score[x-1][-1]
#Example params given. Will be decided by optimizing.
def gen_mlp():
clf = MLPClassifier(alpha=1e-5, hidden_layer_sizes=(7,3), random_state=1, shuffle=True, warm_start=1, verbose=1)
clf.fit(X_train_2008, Y_train_2008)
return clf
# Save Model
mlp = read_make_pkl("saved_objs/mlp.pkl", gen_mlp)
def main(argv):
try:
opts, args = getopt.getopt(argv,"ho:",["output="])
except getopt.GetoptError:
print 'mlpclassifier.py [-o [2008] [2012] [tune]]'
sys.exit(2)
for opt, arg in opts:
if opt == '-h':
print 'mlpclassifer.py [-o [2008] [2012] [tune]]'
sys.exit()
elif opt in ("-o", "--output"):
if (arg == "2008"):
# DEBUG
print("mlp.predict(X_test_2008).shape" +
neighbor = n
print "Number of neighbors: ", neighbor,
print "Training Score: ", train_score[-1], ". Verification Score: ", \
ver_score[-1]
return neighbor
#Example params given. Will be decided by optimizing.
def gen_knearest():
clf = KNeighborsClassifier(n_neighbors=10)
clf.fit(X_train_2008, Y_train_2008)
return clf
# Save Model
knearest = read_make_pkl("saved_objs/knearest.pkl", gen_knearest)
def main(argv):
try:
opts, args = getopt.getopt(argv, "ho:", ["output="])
except getopt.GetoptError:
print 'knearest_neighbor.py [-o [2008] [2012] [tune]]'
sys.exit(2)
for opt, arg in opts:
if opt == '-h':
print 'knearest_neighbor.py [-o [2008] [2012] [tune]]'
sys.exit()
elif opt in ("-o", "--output"):
if (arg == "2008"):
# DEBUG
import os.path
import numpy as np
import pkl_help
from pkl_help import read_make_pkl
from sklearn import preprocessing
import preprocess_help as ph
def get_csv_data(filename):
data = np.genfromtxt(filename, delimiter=",")
return data
train_2008 = read_make_pkl("saved_objs/train_2008.pkl",
lambda: get_csv_data("data/train_2008.csv"),
compress=True)
test_2008 = read_make_pkl("saved_objs/test_2008.pkl",
lambda: get_csv_data("data/test_2008.csv"),
compress=True)
test_2012 = read_make_pkl("saved_objs/test_2012.pkl",
lambda: get_csv_data("data/test_2012.csv"),
compress=True)
##################################################################################
X_train_2008 = read_make_pkl("saved_objs/X_train_2008.pkl",
lambda: ph.remove_header_and_normalize_train(train_2008),
compress=True)
Y_train_2008 = read_make_pkl("saved_objs/Y_train_2008.pkl",
lambda: ph.grab_train_Y(train_2008),
compress=True)
X_ver_2008 = read_make_pkl("saved_objs/X_ver_2008.pkl",
#bst = xgb.train(plst, dtrain, evals=evallist)
# Generate xgb model
def gen_xgb():
evals = [(X_ver_2008, Y_ver_2008)]
model = xgb.XGBRegressor()
model.fit(X_train_2008,
Y_train_2008,
eval_set=evals,
early_stopping_rounds=10,
verbose=True)
return model
xgb_model = read_make_pkl("saved_objs/xgb.pkl", gen_xgb)
def main(argv):
try:
opts, args = getopt.getopt(argv, "ho:", ["output="])
except getopt.GetoptError:
print 'xgb.py [-o [2008] [2012]]'
sys.exit(2)
for opt, arg in opts:
if opt == '-h':
print 'xgb.py [-o [2008] [2012]]'
sys.exit()
elif opt in ("-o", "--output"):
if (arg == "2008"):
preds = round_predictions(xgb_model.predict(X_test_2008))
# Get training data
X_train_2008 = get_pkl("saved_objs/X_train_2008.pkl")
Y_train_2008 = get_pkl("saved_objs/Y_train_2008.pkl")
X_test_2008 = get_pkl("saved_objs/X_test_2008.pkl")
X_test_2012 = get_pkl("saved_objs/X_test_2012.pkl")
# function to generate lasso model
def gen_lasso():
model = LassoCV(cv=10, n_jobs=-1, verbose=True)
model.fit(X_train_2008, Y_train_2008)
return model
# Save model
lasso = read_make_pkl("saved_objs/lasso.pkl", gen_lasso)
def main(argv):
try:
opts, args = getopt.getopt(argv, "ho:", ["output="])
except getopt.GetoptError:
print 'lasso.py [-o [2008] [2012]]'
sys.exit(2)
for opt, arg in opts:
if opt == '-h':
print 'lasso.py [-o [2008] [2012]]'
sys.exit()
elif opt in ("-o", "--output"):
if (arg == "2008"):
preds = round_predictions(lasso.predict(X_test_2008))
# Generate xgb model
def gen_xgb():
evals = [(X_ver_2008, Y_ver_2008)]
model = xgb.XGBRegressor(n_estimators=100000)
#cvresult = xgb.cv(model.get_params(), dtrain, nfold=10,
# early_stopping_rounds=20)
#print ("CVRESULT:\n" + str(cvresult))
model.fit(X_train_2008,
Y_train_2008,
eval_set=evals,
early_stopping_rounds=20,
verbose=True)
return model
xgb_model = read_make_pkl("saved_objs/xgb_100000_estimators.pkl", gen_xgb)
def xgb_preds_2008():
return xgb_model.predict(X_test_2008)
def xgb_preds_2012():
return xgb_model.predict(X_test_2012)
read_make_pkl("saved_objs/xgb_100k_estimators_preds_2008.pkl", xgb_preds_2008)
read_make_pkl("saved_objs/xgb_100k_estimators_preds_2012.pkl", xgb_preds_2012)
def main(argv):
#bst = xgb.train(plst, dtrain, evals=evallist)
# Generate xgb model
def gen_xgb():
evals = [(X_ver_2008, Y_ver_2008)]
model = xgb.XGBRegressor(n_estimators=1000)
model.fit(X_train_2008,
Y_train_2008,
eval_set=evals,
early_stopping_rounds=20,
verbose=True)
return model
xgb_model = read_make_pkl("saved_objs/xgb_1000_estimators.pkl", gen_xgb)
def main(argv):
try:
opts, args = getopt.getopt(argv, "ho:", ["output="])
except getopt.GetoptError:
print 'xgb_1000_estimators.py [-o [2008] [2012]]'
sys.exit(2)
for opt, arg in opts:
if opt == '-h':
print 'xgb_1000_estimators.py [-o [2008] [2012]]'
sys.exit()
elif opt in ("-o", "--output"):
if (arg == "2008"):
preds = round_predictions(xgb_model.predict(X_test_2008))
print "Estimators: ", estimators, ". Max Depth: ", depth, ":"
print "Training Score: ", train_score[estimators-1][-1], ". Verification Score: ", \
ver_score[estimators-1][-1]
return (max_depth, max_est)
#Example params given. Will be decided by optimizing.
def gen_adaboost():
clf = AdaBoostClassifier(DecisionTreeClassifier(max_depth=1),
n_estimators=100)
clf.fit(X_train_2008, Y_train_2008)
return clf
# Save Model
adaboost = read_make_pkl("saved_objs/adaboost.pkl", gen_adaboost)
def main(argv):
try:
opts, args = getopt.getopt(argv, "ho:", ["output="])
except getopt.GetoptError:
print 'adaboost.py [-o [2008] [2012] [tune]]'
sys.exit(2)
for opt, arg in opts:
if opt == '-h':
print 'adaboost.py [-o [2008] [2012] [tune]]'
sys.exit()
elif opt in ("-o", "--output"):
if (arg == "2008"):
# DEBUG
import os.path
import numpy as np
import pkl_help
from pkl_help import read_make_pkl
from sklearn import svm
import preprocess_help as ph
from sklearn.model_selection import train_test_split
from sklearn.model_selection import GridSearchCV
from sklearn.metrics import classification_report
from sklearn.svm import SVC
X_train_2008 = read_make_pkl(
"saved_objs/X_train_2008.pkl",
lambda: ph.remove_header_and_normalize_train(train_2008),
compress=True)
Y_train_2008 = read_make_pkl("saved_objs/Y_train_2008.pkl",
lambda: ph.grab_train_Y(train_2008),
compress=True)
X_ver_2008 = read_make_pkl(
"saved_objs/X_ver_2008.pkl",
lambda: ph.remove_header_and_normalize_ver(train_2008),
compress=True)
Y_ver_2008 = read_make_pkl("saved_objs/Y_ver_2008.pkl",
lambda: ph.grab_ver_Y(train_2008),
compress=True)
params = [{
'kernel': ['rbf'],
'gamma': [1e-3, 1e-4],
'C': [1, 10, 100, 1000]
}, {
print("d1: " + str(model.predict(X_test_2008).shape))
print("d2: " + str(model.coef_.shape))
#
return model
def lin_reg_modified_predict(model, X):
preds = model.predict(X).reshape(-1, 1)
mpreds = round_predictions(preds)
# Debug
print("mpreds.shape: " + str(mpreds.shape))
return mpreds
# Save model
lin_reg_model = read_make_pkl("saved_objs/lin_reg.pkl", gen_lin_reg_model)
def main(argv):
try:
opts, args = getopt.getopt(argv, "ho:", ["output="])
except getopt.GetoptError:
print 'linear_regression.py [-o [2008] [2012]]'
sys.exit(2)
for opt, arg in opts:
if opt == '-h':
print 'linear_regression.py [-o [2008] [2012]]'
sys.exit()
elif opt in ("-o", "--output"):
if (arg == "2008"):
# DEBUG
model = RandomForestClassifier(max_depth=15,
n_estimators=30,
max_features='auto')
model.fit(X_train_2008, Y_train_2008)
return model
def rand_forest_modified_predict(model, Y):
preds = model.predict(Y).reshape(-1, 1)
# Debug
print("preds.shape: " + str(preds.shape))
return preds
# Save Model
rand_forest_model = read_make_pkl("saved_objs/rand_forest.pkl",
gen_rand_forest)
def main(argv):
try:
opts, args = getopt.getopt(argv, "ho:", ["output="])
except getopt.GetoptError:
print 'random_forest.py [-o [2008] [2012] [graphs]]'
sys.exit(2)
for opt, arg in opts:
if opt == '-h':
print 'random_forest.py [-o [2008] [2012] [graphs]]'
sys.exit()
elif opt in ("-o", "--output"):
if (arg == "2008"):
# DEBUG
# Get training data
X_train_2008 = get_pkl("saved_objs/X_train_2008.pkl")
Y_train_2008 = get_pkl("saved_objs/Y_train_2008.pkl")
X_test_2008 = get_pkl("saved_objs/X_test_2008.pkl")
X_test_2012 = get_pkl("saved_objs/X_test_2012.pkl")
# function to generate ridge model
def gen_ridge():
model = RidgeCV()
model.fit(X_train_2008, Y_train_2008)
return model
# Save model
ridge = read_make_pkl("saved_objs/ridge.pkl", gen_ridge)
def main(argv):
try:
opts, args = getopt.getopt(argv, "ho:", ["output="])
except getopt.GetoptError:
print 'ridge.py [-o [2008] [2012]]'
sys.exit(2)
for opt, arg in opts:
if opt == '-h':
print 'ridge.py [-o [2008] [2012]]'
sys.exit()
elif opt in ("-o", "--output"):
if (arg == "2008"):
preds = round_predictions(ridge.predict(X_test_2008))
def gen_voting():
print "Making Voting Classifier"
clf = VotingClassifier(estimators=[('ridge', ridge), ('mlp', mlp),
('rand_forest', rand_forest),
('adaboost', adaboost)],
voting='hard')
print "Training Voting Classifier"
clf.fit(X_train_2008, Y_train_2008)
print "Scoring Voting Classifier"
print "Training Score: ", clf.score(X_train_2008, Y_train_2008)
print "Ver Score: ", clf.score(X_ver, Y_ver)
return clf
# Save Model
voting = read_make_pkl("saved_objs/voting.pkl", gen_voting)
def main(argv):
try:
opts, args = getopt.getopt(argv, "ho:", ["output="])
except getopt.GetoptError:
print 'voting.py [-o [2008] [2012]]'
sys.exit(2)
for opt, arg in opts:
if opt == '-h':
print 'voting.py [-o [2008] [2012]]'
sys.exit()
elif opt in ("-o", "--output"):
if (arg == "2008"):
# DEBUG
