"""
Creates a box plot of the features on the x axis against the respective values on the y axis
The shape and layout varies with plt.boxplot()
"""
import matplotlib.pyplot as plt
from HelpMethods import helpers, given
DATA_PATH = ""
SAVE_PATH = ""
yb, input_data, ids = given.load_csv_data(DATA_PATH)
cleaned = helpers.clean_matrix(input_data)
stand_data = helpers.standardize(cleaned)
def box_plot_features_values(data, save):
plt.figure()
plt.boxplot(data, 0, '')
plt.xlabel("Feature")
plt.ylabel("Value")
plt.savefig(save + str("Values") + "-" + str("Features") + " plot.png",
dpi=200)
plt.clf()
box_plot_features_values(stand_data, SAVE_PATH)
from HelpMethods import helpers, given from Implementations import implementations # Our best submission, least squares using a polynomial expansion of 6. TRAIN_DATA_PATH = "" TEST_DATA_PATH = "" PREDICTION_NAME = "Least Squares with degree 6." degree = 6 # load training data y_train, x_train, _ = given.load_csv_data(TRAIN_DATA_PATH) # build the polynomial for training training = helpers.build_poly(x_train, degree) # calculating weights weights, mse_train = implementations.least_squares(y_train, training) # load test data y_test, x_test, ids_test = helpers.load_csv_data(TEST_DATA_PATH) # build the polynomial for testing test = helpers.build_poly(x_test, degree) prediction = given.predict_labels(weights, test) given.create_csv_submission(ids_test, prediction, PREDICTION_NAME)
"""
Example of how we used split data to search for optimal hyper-parameters using a grid search.
"""
from HelpMethods import helpers, given
DATA_PATH = ""
SPLIT_RATIO = 0.8
# load the data
y_train, x_train, ids = given.load_csv_data(DATA_PATH)
#split the data with given ratio
train_x, train_y, test_x, test_y = helpers.split_data(x_train, y_train,
SPLIT_RATIO)
# standardize the train and test data
stand_train = helpers.standardize(train_x)
stand_test = helpers.standardize(test_x)
#Initializing lists of hyper-parameters we want to test
degrees = []
hyper_params = []
#Initializing decision variables
best_rmse = 1000000
best_degree = -1
best_hyperparam = -1
# grid search for best combination of hyperparameters. Can be extended with more for loops.
# change "REGRESSION_METHOD" with desired method.