def mnb():
x, y = process.read_data("train.csv", 1, 1)
x_test, y_test = process.read_data("test.csv", 1, 0)
y_train = y.T
mnb = MultinomialNB(alpha=19.6, fit_prior=False)
mnb.fit(x, y_train)
return mnb, x_test
def lr():
x, y = process.read_data("train.csv", 1, 1)
x_train_preprocessed = preprocessing.scale(x)
x_test, y_test = process.read_data("test.csv", 1, 0)
x_test_preprocessed = preprocessing.scale(x_test)
y_train = y.T
lr = LogisticRegression(C=0.05,
random_state=1,
solver='saga',
multi_class='multinomial',
max_iter=800)
lr.fit(x_train_preprocessed, y_train)
return lr, x_test_preprocessed
def marker_map(self):
a = all_data()
df = read_data()
html_dom_dict = self.gen_html_dom()
html_body = self.gen_html_body()
center_lat_prime = 41.7943 #df['Latitude'].mean()
center_lon_prime = -87.5907# df['Longitude'].mean()
#41.7943° N, 87.5907° W
print (center_lat_prime, center_lon_prime)
colors = html_dom_dict["marker_map_colors"]
map_data =Map.process_data()
points = Map.create_map_points(html_dom_dict, map_data)
return html_body.format(
map_points=points,#map_points_prime,
center_lat=center_lat_prime,
center_lon=center_lon_prime,
colors = colors,
style=html_dom_dict["style"])
def process_data():
df = read_data()
map_data = [[str(y) for y in x[1]] for x in df.iterrows()]
d = {}
k = {}
for street, number, lat, lon in map_data:
num = int(number)
if street in d:
if num > d[street]:
d[street] = num
k[street] = [street, num, lat, lon]
else:
d[street] = num
k[street] = [street, num, lat, lon]
sorted_map_data = sorted(map_data, key = lambda x: x[1])
return sorted_map_data
def svm_train_cv():
x, y = process.read_data("train.csv", 1, 1)
sscaler = StandardScaler()
sscaler.fit(x)
x_train_preprocessed = sscaler.transform(x)
y_train = y.T
best_score = 0.0
num = 0
for coef0 in [0.1, 0.12, 0.14, 0.16]:
svm_rbf = svm.SVC(C=0.90,
cache_size=500,
kernel='sigmoid',
gamma='scale',
coef0=coef0)
num = num + 1
scores = cross_val_score(svm_rbf, x_train_preprocessed, y_train, cv=3)
score = scores.mean()
print("Iteration time:{}th".format(num))
print("Current score on validation set:{:.9f}".format(score))
print("Current parameters:{:.2f}".format(coef0))
if score > best_score:
best_score = score
best_parameters = {"coef0": coef0}
print("Best score on validation set:{:.9f}".format(best_score))
print("Best parameters:{}".format(best_parameters))
def main():
syn_symbols = {SYNCHROTRONS[s][0]: s for s in OUTPUT_SYN}
data = {
s: {y: [0] * days_in_year(y)
for y in OUTPUT_YEARS}
for s in ALL_OUTPUT
}
for items in read_data():
syn = SYNCHROTRONS.get(items[0][11])
name = None
if items[0][8] == 'ELECTRON MICROSCOPY':
# No stats for CryoEM. In most of entries _em_imaging.date is null.
name = None
elif syn:
name = syn_symbols.get(syn[0])
elif items[0][9] in ('SEALED TUBE', 'ROTATING ANODE'):
name = 'home'
if not name:
continue
coll = parse_date(items[0][19])
if not coll or coll.year not in OUTPUT_YEARS:
continue
day = coll.timetuple().tm_yday
data[name][coll.year][day - 1] += 1
# summary to stderr
for syn in ALL_OUTPUT:
totals = [sum(data[syn][year]) for year in OUTPUT_YEARS]
print('%-10.10s %-17s total: %4d' % (syn, totals, sum(totals)),
file=sys.stderr)
# JSON to stdout
print_data(data)
def svm_train():
x, y = process.read_data("train.csv", 1, 1)
print(x)
print(y)
x_test, y_test = process.read_data("test.csv", 1, 0)
sscaler = StandardScaler()
sscaler.fit(x)
x_train_preprocessed = sscaler.transform(x)
x_test_preprocessed = sscaler.transform(x_test)
y_train = y.T
svm_rbf = svm.SVC(C=0.90,
cache_size=500,
kernel='sigmoid',
gamma='auto',
coef0=0.1)
svm_rbf.fit(x_train_preprocessed, y_train)
return svm_rbf, x_test_preprocessed
def read_text(f):
"""
Reads text from file
f<string> -> string
"""
if Globals.override:
from process import read_data
return read_data()
with open(f, 'r') as infile:
return infile.read().lower()
def mlp_train():
x, y = process.read_data("train.csv", 1, 1)
sscaler = StandardScaler()
sscaler.fit(x)
x_train_preprocessed = sscaler.transform(x)
y_train = y.T
mlp = MLPClassifier(activation='logistic',
solver='sgd',
alpha=1e-5,
batch_size=200,
hidden_layer_sizes=(50, 50),
random_state=1,
learning_rate='adaptive',
max_iter=600)
scores_mlp = cross_val_score(mlp, x_train_preprocessed, y_train, cv=3)
print(scores_mlp)
print("scores_mlp Accuracy: %0.9f (+/- %0.9f)" %
(scores_mlp.mean(), scores_mlp.std() * 2))
def cnb_cv():
x, y = process.read_data("train.csv", 1, 1)
#x_train_preprocessed = preprocessing.scale(x)
y_train = y.T
best_score = 0.0
num = 0
for alpha in [0.00001, 0.0001, 0.001, 0.01, 0.1, 1]:
num = num + 1
cnb = ComplementNB(alpha=alpha, fit_prior=False)
scores = cross_val_score(cnb, x, y_train, cv=4)
score = scores.mean()
print("Iteration time:{}th".format(num))
print("Current score on validation set:{:.9f}".format(score))
print("Current parameters:{:.2f}".format(alpha))
if score > best_score:
best_score = score
best_parameters = {"alpha": alpha}
print("Best score on validation set:{:.9f}".format(best_score))
print("Best parameters:{}".format(best_parameters))
def lr_cv():
x, y = process.read_data("train.csv", 1, 1)
x_train_preprocessed = preprocessing.scale(x)
y_train = y.T
best_score = 0.0
num = 0
for C in [0.03, 0.05, 0.07]:
num = num + 1
lr = LogisticRegression(C=C,
random_state=1,
solver='saga',
multi_class='multinomial',
max_iter=1000,
penalty='l2')
scores = cross_val_score(lr, x_train_preprocessed, y_train, cv=3)
score = scores.mean()
print("Iteration time:{}th".format(num))
print("Current score on validation set:{:.9f}".format(score))
print("Current parameters:{:.4f}".format(C))
if score > best_score:
best_score = score
best_parameters = {"C": C}
print("Best score on validation set:{:.9f}".format(best_score))
print("Best parameters:{}".format(best_parameters))