def __init__(self, image_file, label_file, patch_size=(101, 101)):
self.input = prepared_dataset_image(image_file, border=patch_size)
self.image_size = image_size(prepared_dataset_image(image_file))
self.patch_size = patch_size
width, height = self.image_size
self.output = np.zeros((height, width))
self.verbose = TT.verbose
for (col, row, p) in load_csv(label_file):
self.output[row, col] = 1.0
def sim3():
filename = '../dataset/dataset3.csv'
dataset = load_csv(filename)
dataset = dataset_to_float(dataset)
plot_2D_dataset(dataset, "Simulation n.3")
tree = build_tree(dataset, 5, 1)
print("-" * 10 + " Sim.3 TREE " + "-" * 10)
print_tree(tree.root, ['x', 'y'])
print_tree_separating_2D(tree.root)
show_plot()
def main():
start_season = 1995
end_season = 2020
regular_season = True
# Only need to update drafts once a year
if False:
drafts = util.get_drafts(start_season, end_season) # Takes a loonnnggg time to run
util.save_csv("drafts.csv", drafts)
else:
drafts = util.load_csv("drafts.csv")
if True:
players = get_career_stats(start_season, end_season, regular_season)
util.save_csv("players.csv", players)
rosters = util.get_rosters()
util.save_csv("rosters.csv", rosters)
else:
players = util.load_csv("players.csv")
rosters = util.load_csv("rosters.csv")
drafts = update_team_names(drafts)
drafts = drafts.sort_values(["team.name", "year", "round"], ascending=[1, 0, 1])
# Merge all data into one dataframe
drafts['name_lower'] = drafts['prospect.fullName'].str.lower()
players['name_lower'] = players['playerName'].str.lower()
rosters['name_lower'] = rosters['fullName'].str.lower()
draft_data = pd.merge(drafts, players, how="left", on="name_lower", sort=False, suffixes=("", "_x"))
draft_data = pd.merge(draft_data, rosters, how="left", on="name_lower", sort=False, suffixes=("", "_y"))
# Update positions and set statuses for each filter in the visuzalization. Then get rid of unneeded columns.
draft_data = set_statuses(draft_data)
draft_data = clean_data(draft_data)
draft_data = reduce_columns(draft_data)
util.save_csv("draft_data.csv", draft_data)
def positive(self):
if hasattr(self, '_positive'):
return self._positive, self._positive_size
TT.debug("Collecting positive samples.")
self._positive = {}
self._positive_size = 0
self._positive_expanded = {}
for data_file, label_file in self.files:
labels = load_csv(os.path.join(self.root_path, label_file))
self._positive[data_file] = labels
self._positive_size += len(labels)
self._positive_expanded[data_file] = {}
for col, row, p in labels:
self._positive_expanded[data_file][index_at_pixel(col=col, row=row, size=self.image_size)] = p
TT.debug("Found", self._positive_size, "positive samples.")
return self.positive
def sim4():
filename = '../dataset/dataset3.csv'
dataset = load_csv(filename)
dataset = dataset_to_float(dataset)
train_accuracy = list()
nodes_numbers = list()
for i in range(1, 21):
tree_size, predicted, expected = decision_tree_prediction_and_size(
dataset, dataset, i, 1)
acc = accuracy_metric(expected, predicted)
train_accuracy.append(acc)
nodes_numbers.append(tree_size)
x = range(1, 21)
plt.figure("Simulation n.4")
line1, = plt.plot(x, train_accuracy, 'r', label='Train accuracy')
line2, = plt.plot(x, nodes_numbers, 'g', label='Tree complexity(# nodes)')
plt.legend(handles=[line1, line2], loc=4)
plt.xticks(x)
plt.xlabel('Maximum Tree Depth')
plt.draw()
def parkinson_main(n_folds_outer_cross_val, n_folds_inner_cross_val, max_depth, min_size):
if n_folds_outer_cross_val < 2 or n_folds_inner_cross_val < 2:
raise ValueError("Illegal value parameter")
filename = '../dataset/parkinson_recording_data.csv'
dataset = load_csv(filename)
dataset = dataset_to_float(dataset)
folds = cross_validation_split(dataset,n_folds_outer_cross_val)
scores = list()
outer_fold_number = 0
# Outer k-fold cross validation
for fold in folds:
outer_fold_number += 1
# Prepare train and test set
train_set = list(folds)
train_set.remove(fold)
train_set = sum(train_set, [])
test_set = list()
for row in fold:
row_copy = list(row)
test_set.append(row_copy)
row_copy[-1] = None
# Inner k-fold cross validation ( grid search )
best_couple, inner_accuracy = two_grid_search_with_accuracy_return(decision_tree, n_folds_inner_cross_val, train_set, max_depth, min_size)
# Evaluate results on outer cross validation test set
predictions = decision_tree(train_set, test_set, best_couple[0], best_couple[1])
actual = [row[-1] for row in fold]
outer_accuracy = accuracy_metric(actual, predictions)
print("-" * 10 + " Outer Fold n. " + str(outer_fold_number) + " " + "-" * 10)
print("Best params selected by inner cross validation (max_depth,min_size): "+str(best_couple[0])+" "+str(best_couple[1]))
print("Best params mean accuracy in the inner cross validation: " + str(inner_accuracy))
print("Best params accuracy in the outer cross validation: " + str(outer_accuracy))
scores.append(outer_accuracy)
print("-" * 10 + " Final Results " + " " + "-" * 10)
print("Total Accuracy mean: " + str(mean(scores)))
print("Total Accuracy std dev: " + str(stdev(scores)))
return scores
def banknote_main(tr_percentages, number_repetitions, n_folds_2grid_search,
max_depth, min_size):
if number_repetitions < 2 or n_folds_2grid_search < 2:
raise ValueError("Illegal value parameter")
filename = '../dataset/data_banknote_authentication.csv'
dataset = load_csv(filename)
dataset = dataset_to_float(dataset)
mean_accuracies = list()
std_devs = list()
# For each percentage of training split
for percentage in tr_percentages:
accuracies = list()
# Repeat number_repetions times the random split and test validation for each split
for run in range(1, number_repetitions + 1):
train, test = random_training_test_split(dataset, percentage)
result = two_grid_search(decision_tree, n_folds_2grid_search,
train, max_depth, min_size)
predictions = decision_tree(train, test, result[0], result[1])
actual = [row[-1] for row in test]
accuracy = accuracy_metric(actual, predictions)
accuracies.append(accuracy)
print("-" * 10 + " training split %" + str(percentage) + " " +
"-" * 10)
print("Accuracies of training split %" + str(percentage) + " : " +
str(accuracies))
mean_acc = mean(accuracies)
std_dev = stdev(accuracies)
print("Accuracy mean: " + str(mean_acc))
print("Accuracy std dev: " + str(std_dev))
mean_accuracies.append(mean_acc)
std_devs.append(std_dev)
plt.figure("BankNote dataset")
plot_results(tr_percentages, mean_accuracies, std_devs)
plt.show()