temperature(a) + [1, 0, 0],
temperature(0) + [0, 1, 0],
temperature(0) + [0, 0, 1]
])
y.append([[b / 10.0], [a / 10.0], [((a + b) % 10) / 10.0],
[((a + b) // 10) / 10.0]])
x = np.array(x)
y = np.array(y)
test_x = np.array(test_x)
test_y = np.array(test_y)
model = LSTMModel(4, [512, 512], 12, 1)
model.train(x, y, test_x, test_y, 10, 5000)
results = model.predict(x)
count = 0
for index, result in enumerate(results):
left = round(result[2][0], 1)
right = round(result[3][0], 1)
leftTarget = round(y[index][2][0], 1)
rightTarget = round(y[index][3][0], 1)
if left == leftTarget and right == rightTarget:
count += 1
test_count = 0
test_results = model.predict(test_x)
print("Test dataset has {} samples.".format(*x_test.shape))
# print(x_test[:3])
# print(y_test[:3])
# Build & train model
model = LSTMModel(timesteps, hidden_neurons).build()
print("Fitting the model...")
history = model.fit(x_train, y_train,
batch_size=batchsize, epochs=epochs, validation_data=(x_val, y_val))
# Output training history to csv
history_df = pd.DataFrame(history.history)
history_df.to_csv("training_history.csv")
print("Predicting...")
result = model.predict(x_test)
predicted = pd.DataFrame(result)
predicted.columns = ['predicted_nikkei']
predicted['actual_nikkei'] = y_test
print("Completed Prediction.")
print(predicted.shape)
print(predicted[:10])
# Output to csv
predicted.to_csv("predicted.csv")
# Evaluate
evaluation_score = model.evaluate(x_test, y_test, batch_size=batchsize, verbose=1)
print("Evaluation score is {}".format(evaluation_score))
# Evaluate benchmark
[
noisyTrain,
noisySymbolsTrain,
targetsTrain,
noisyValidation,
noisySymbolsValidation,
targetsValidation,
noisyTest,
noisySymbolsTest,
targetsTest
] = dataLoader.getSequencialDataWithDontCare(0, 2, 2)
model = LSTMModel(2, [512, 512, 512], 1568, 20)
model.train(noisyTrain, targetsTrain, noisyValidation, targetsValidation, 100, 100)
result = model.evaluate(noisyTest, targetsTest)
predictions = model.predict(noisyTest)
count = 0
for index in range(len(predictions)):
a = utils.argmax(predictions[index][:10])
b = utils.argmax(predictions[index][10:])
predicted = a * 10 + b
a = utils.argmax(targetsTest[index][:10])
b = utils.argmax(targetsTest[index][10:])
actual = a * 10 + b
if actual == predicted:
count += 1
print("ACCURACY: " + str(result[1]))
])
y.append([
one_hot(c),
one_hot(b),
one_hot(a),
one_hot((a + b + c) % 10),
one_hot((a + b + c) // 10)
])
x = np.array(x)
y = np.array(y)
model = LSTMModel(5, [10, 10], 12, 12)
model.train(x, y, x, y, 10, 5000)
results = model.predict(x)
activations_0 = model.getActivations(0, x)
activations_1 = model.getActivations(1, x)
count = 0
for index, result in enumerate(results):
left = argmax(result[3])
right = argmax(result[4])
leftTarget = argmax(y[index][3])
rightTarget = argmax(y[index][4])
if left == leftTarget and right == rightTarget:
count += 1
for i in range(10):
def main():
"""Program execution starts here."""
intersections = read_intersections("intersections.data")
intersection_list = []
for i in range(len(intersections)):
intersection_list.append(intersections[i]['name'].strip(' '))
#sel_intersection = intersections[0]['name']
sel_intersection = "4589"
df = get_dataset(sel_intersection, intersections)
x_train, x_test, y_train, y_test = preprocessing(df)
model_file = search(intersections, 'name', sel_intersection)
main_menu = [
"CAPSTONE TRAFFIC PREDICTION",
"Select Intersection",
"List of Intersections",
"Train Intersection",
"Test Intersection (Accuracy)",
"Route Check"
]
train_menu = [
"Train Mode:",
"Train from scratch w/ events",
"Train from file w/ events",
"Train from scratch",
"Train from file",
]
route_check_menu = [
"Train Mode:",
"Train from scratch w/ events",
"Train from file w/ events",
"Train from scratch",
"Train from file",
]
while True:
print("Currently Selected Intersection:", sel_intersection)
choice = menu.do_menu(main_menu)
model = LSTMModel(x_train.shape[1], y_train.shape[1])
if choice is None:
return # Exit main() (and program).
if choice == 1:
# Select Intersection
temp_menu = ["Please Select a New Intersection"]
for line in intersections:
option = line["name"] + ": " + line["street"]
temp_menu.append(option)
choice = menu.do_menu(temp_menu)
if choice is not None:
sel_intersection = intersections[choice - 1]['name']
print(sel_intersection, "set as current intersection.")
df = get_dataset(sel_intersection, intersections)
x_train, x_test, y_train, y_test = preprocessing(df)
model_file = search(intersections, 'name', sel_intersection)
if model_file is not None:
model_file = intersections[model_file]['model']
#try:
model.load_network('./model/'+str(sel_intersection)+'.hdf')
#except:
# print("File for loading model is not found! Creating a new model from scratch")
# model.init_network(hidden_size=50)
elif choice == 2:
# List Intersections
print_intersections(intersections)
elif choice == 3:
model = LSTMModel(x_train.shape[1], y_train.shape[1])
# Train Intersections
# TODO test each option
choice = menu.do_menu(train_menu)
if choice == 1:
x_train, x_test, y_train, y_test = preprocessing(df, events=True)
model = LSTMModel(x_train.shape[1], y_train.shape[1])
intersection_idx = search(intersections, 'name', sel_intersection)
model_file = "model/" + sel_intersection + ".hdf"
#if os.path.exists(model_file):
#os.remove(model_file)
model.init_network(hidden_size=50)
elif choice == 2:
x_train, x_test, y_train, y_test = preprocessing(df, events=True)
model = LSTMModel(x_train.shape[1], y_train.shape[1])
intersection_idx = search(intersections, 'name', sel_intersection)
model_file = "model/" + sel_intersection + ".hdf"
if os.path.exists(model_file):
model.load_network(model_file)
else:
print("Model does not exist, starting from scratch")
model.init_network(hidden_size=50)
elif choice == 3:
x_train, x_test, y_train, y_test = preprocessing(df, events=False)
model = LSTMModel(x_train.shape[1], y_train.shape[1])
intersection_idx = search(intersections, 'name', sel_intersection)
model_file = "model/" + sel_intersection + ".hdf"
if os.path.exists(model_file):
os.remove(model_file)
model.init_network(hidden_size=50)
elif choice == 4:
x_train, x_test, y_train, y_test = preprocessing(df, events=False)
model = LSTMModel(x_train.shape[1], y_train.shape[1])
intersection_idx = search(intersections, 'name', sel_intersection)
model_file = "model/" + sel_intersection + ".hdf"
if os.path.exists(model_file):
model.load_network(model_file)
else:
print("Model does not exist, starting from scratch")
model.init_network(hidden_size=50)
try:
e = int(input("Enter Epochs to train for (Default=50): "))
model.epochs = e
except ValueError:
print("Invalid number entered, using default value")
model.train(x_train, y_train, './'+ model_file)
intersections[intersection_idx]['model'] = model_file
save_intersections(intersections, "intersections.data")
elif choice == 4:
# Test Intersections
model_file = "model/" + sel_intersection + ".hdf"
model = LSTMModel(x_train.shape[1], y_train.shape[1])
if os.path.exists(model_file):
model.load_network(model_file)
test_output = model.get_accuracy(x_test, y_test)
N, S, E, W = confusion_matrix(test_output, y_test, True)
else:
print("Please train intersection first")
elif choice == 5:
model = LSTMModel(x_train.shape[1], y_train.shape[1], intersection_list)
# Route Check
x_data = []
day_week = [1,2,3,4,5,6,7]
season=[1,2,3,4]
time_str = ''
flag = 0
x_data = []
peak = 0
while flag == 0:
num_inter = input("Please enter intersection number: ")
if num_inter in intersection_list:
flag = 1
num_inter = int(num_inter)
#x_data.append(int(num_inter))
else:
print("Intersection not found!")
flag = 0
while flag == 0:
week = [0,0,0,0,0,0,0]
num_day = input("Please enter the day of the week:\nOptions:\n1:Sunday\n2:Monday\n3:Tuesday\n4:Wednesday\n5:Thursday\n6:Friay\n7:Saturday\n")
num_day = int(num_day)
if num_day in day_week:
flag = 1
week[num_day-1] = 1
x_data = x_data + week
else:
print("Day of the week not found!")
flag = 0
while flag == 0:
time_day = input("Please enter the time of the day (ex. 17:30): ")
temp = time_day.split(':')
if len(temp) == 2:
hour = int(temp[0]) * 60
if hour > 9 and hour < 17:
peak = 1
time = hour + int(temp[1])
time_d = float(time) / float(1440)
if time_d > 1.0:
print("Please enter time in the proper format!")
else:
x_data.append(time_d)
flag = 1
else:
print("Please enter time in the proper format!")
flag = 0
while flag == 0:
seasons = [0,0,0,0]
season_input = input("Please enter the season:\n1:Summer\n2:Fall\n3:Winter\n4:Spring\n")
season_input = int(season_input)
if season_input in season:
flag = 1
seasons[season_input-1] = 1
x_data = x_data + seasons
else:
print("Season not found!")
# add [0,0] for events
x_data.append(peak)
x_data = x_data + [0,0]
x_test = np.array([x_data])
x_test = np.reshape(x_test, (x_test.shape[0], x_test.shape[1], 1))
res = model.predict(x_test, num_inter)
print("Prediction: " + str(res))