def draw_wr_hist(save=False):
# Load the winrate matrix
wr_matrix = load_data('Winrate')
winrate_count = [round(rate[0] * 100, 2) for rate in wr_matrix]
# Setup the x-axis values
index = np.arange(len(heroes_list))
plt.figure(figsize=(20, 30), dpi=100)
plt.barh(index, winrate_count, height=0.5, color='peru')
plt.ylabel('Heroes', fontsize=30, labelpad=20)
plt.xlabel('Win Rate', fontsize=30, labelpad=20)
plt.xlim(35,60, 1)
plt.yticks(index, heroes_list, fontsize=12, rotation='horizontal', style='italic')
plt.margins(y=0.01)
plt.title('Win Rate of Each Hero', fontsize=50)
ax = plt.gca()
# Moidfy the title
ax.title.set_position([.5, 1.01])
# Make the count values to show next to the bars
for i, v in enumerate(winrate_count):
ax.text(v + 0.5, i - 0.3, str(v), color='darkred')
if not save:
plt.show()
else:
plt.savefig('hero_win_rate.png')
def draw_heatmap(save=False):
syn_matrix, coun_matrix = load_data('SynCoun')
matrices = {'syn_matrix': syn_matrix, 'coun_matrix': coun_matrix}
# Draw syn matrix
for key in matrices:
annotation = np.round(matrices[key], 2)
sns.set(font_scale=2)
fig, ax = plt.subplots(figsize=(60, 60))
# Set pallette for colorbar
if key == 'syn_matrix':
cmap = "YlGnBu"
else:
cmap = "OrRd"
ax = sns.heatmap(matrices[key], square=True,
linewidths=.1, cmap=cmap, cbar_kws={"shrink": .80},
xticklabels=heroes_list, yticklabels=heroes_list, annot_kws={"size": 8})
plt.xlabel('Hero 2', fontsize=150, labelpad=40)
plt.ylabel('Hero 1', fontsize=150, labelpad=40)
# Modify colorbar labels
cax = plt.gcf().axes[-1]
cax.tick_params(labelsize=50)
if not save:
plt.show()
else:
plt.savefig(key + '.png')
def draw_heroes_hist(save=False):
X_matrix, y_matrix, X_train, X_test, y_train, y_test = load_data('Base')
# A list of the total amount of matches for hero ID i which correspond to the index of this list
heroes_count = []
for i in range(TOTAL_NUM_HEROES):
heroes_count.append(np.count_nonzero(X_matrix[:, i]) / X_matrix.shape[0])
# Setup the x-axis values
index = np.arange(len(heroes_list))
plt.figure(figsize=(20, 30), dpi=100)
plt.barh(index, heroes_count, height=0.5)
plt.ylabel('Heroes', fontsize=30, labelpad=20)
plt.xlabel('Percentage', fontsize=30, labelpad=20)
plt.yticks(index, heroes_list, fontsize=12, rotation='horizontal', style='italic')
plt.margins(y=0.01)
plt.title('Percentage of matches where a hero is present', fontsize=50)
ax = plt.gca()
# Moidfy the title
ax.title.set_position([.5, 1.01])
# Make the count values to show next to the bars
for i, v in enumerate(heroes_count):
percent = round(v * 100, 2)
ax.text(v + 0.01, i - 0.3, str(percent), color='blue')
if not save:
plt.show()
else:
plt.savefig('hero_num_matches.png')
def draw_rf_trees():
X_matrix, y_matrix, X_train, X_test, y_train, y_test = load_data('Base')
number_trees = np.linspace(100, 1000, 10, dtype=int)
results = []
for tree in tqdm(number_trees):
rf = RandomForestClassifier(random_state=1411, n_estimators=tree, warm_start=True)
rf.fit(X_train, y_train)
results.append(rf.score(X_test, y_test))
plt.plot(number_trees, results)
plt.xlabel('Number of trees')
plt.ylabel('Test Accuracy')
plt.savefig('RF tree.png')
def web_heatmap():
syn_matrix, coun_matrix = load_data('SynCoun')
matrices = {'syn_matrix': syn_matrix, 'coun_matrix': coun_matrix}
for key in matrices:
if key == 'syn_matrix':
cmap = "YlGnBu"
title = 'Heatmap of synergistic relationship between heroes'
else:
cmap = "YlOrRd"
title = 'Heatmap of antagonistic relationship of Hero 1 (y) against Hero 2 (x)'
trace = [go.Heatmap(z=matrices[key], x=heroes_list, y=heroes_list, colorscale=cmap)]
layout = go.Layout(title=title, xaxis=dict(title='Hero 2'),
yaxis=dict(title='Hero 1', autorange='reversed'))
fig = go.Figure(data=trace, layout=layout)
py.iplot(fig, filename=key + "_heatmap")
except ValueError:
print("Invalid input")
exit()
else:
data_name = which_data
if which_data == 1:
data_name = 'Base'
elif which_data == 2:
data_name = 'Perf'
else:
print("Invalid input")
exit()
# Load data file and define matrices
try:
X_matrix, y_matrix, X_train, X_test, y_train, y_test = load_data(
data_name)
except:
print('Could not locate data, run "preprocess_data.py" first.')
exit()
else:
while True:
try:
user_input = int(
input(
"Lists of options to input:\n1. Fill syn and coun matrices and save them \n"
"2. Load the %s dataset and append the synergy counter matrices\n3. Exit\n"
% data_name))
except ValueError:
print("Invalid input")
continue
print(
'ROC AUC score of logistic regression classifier on test set: {:.2f} \n'
.format(roc_auc_score(y_test, y_pred)))
print("Classification Report:\n")
print(classification_report(y_test, y_pred))
print("Confusion Matrix: \n")
print(confusion_matrix(y_test, y_pred))
return clf
if __name__ == '__main__':
np.random.seed(1411)
dataset = input("What dataset to use: \n")
number_cv = int(input("How many cv folds: \n"))
X_matrix, y_matrix, X_train, X_test, y_train, y_test = load_data(
type=dataset, scale=False)
# Output results to txt
text_file = os.path.join(BASE_DIR, TRAINED_MODEL_DIR,
dataset + 'LogModel.txt')
sys.stdout = open(text_file, "w")
print("Fitting model: \n")
# Fitting basic model
logreg = fit_model()
predict(logreg, X_test, y_test)
cross_validation(logreg, X_matrix, y_matrix, cv=number_cv)
print()
print("#" * 100)
tmp_threshold = model.score(X_test_new, y_test_new)
print("Number of matches where " + text + " are removed: %s" % X_new.shape[0])
print("Test accuracy is: %s" % tmp_threshold)
print("#" * 100)
print()
return tmp_threshold
if __name__ == '__main__':
np.random.seed(1411)
model = LogisticRegression()
X_matrix, y_matrix, X_train, X_test, y_train, y_test = load_data('Base')
model.fit(X_train, y_train)
highest_coefs = np.argsort(model.coef_[0])[::-1]
heroes_list = get_heroes_names()
# Base threshold
threshold = model.score(X_test, y_test)
wr_matrix = load_data('Winrate')
print("Test accuracy of base: %s" % threshold)
print("#" * 100)
print()
# Setup simluation test information
sim_count = 1
num_simulations = 100
def client_exit(self):
exit()
class ChooseModels:
def __init__(self, master):
Frame.__init__(master)
if __name__ == '__main__':
# Get the list of heroes and list of core and support heroes
heroes_list, heroes_dict, core_list, support_list = get_heroes()
get_heroes_img()
# Load features data
syn_matrix, coun_matrix = load_data('SynCoun')
wr_matrix = load_data('Winrate')
time_matrix, pick_matrix = load_data('TimePickWr')
root = Tk()
root.wm_state('zoomed')
rows, columns = 0, 0
while rows < 40:
root.grid_rowconfigure(rows, weight=20)
rows += 1
while columns < 80:
root.grid_columnconfigure(columns, weight=20)
columns += 1
team_wr.append(hero_pick_wr)
return np.mean(team_wr)
else:
return 0
if __name__ == '__main__':
# Create new matrices
win_pick_matrix = np.zeros((TOTAL_NUM_HEROES, 5))
total_pick_matrix = np.zeros((TOTAL_NUM_HEROES, 5))
win_time_matrix = np.zeros((TOTAL_NUM_HEROES, 5))
total_time_matrix = np.zeros((TOTAL_NUM_HEROES, 5))
# Load the base data with match duration, pick order information
X_matrix, y_matrix, X_train, X_test, y_train, y_test = load_data('TimePickBase')
X_matrix2, y_matrix2, X_train2, X_test2, y_train2, y_test2 = load_data('TimePickPerfBase')
# Filling the matrices
pick_matrix = fill_pick_matrix()
print("#" * 100)
time_matrix = fill_time_matrix()
print("#" * 100)
# Add 4 new columns for the pick order and match duration feature,, 2 for each team
extra_columns = np.zeros((len(X_matrix), 4))
X_matrix = np.append(X_matrix, extra_columns, axis=1)
X_matrix2 = np.append(X_matrix2, extra_columns, axis=1)
for match in range(len(X_matrix)):
# Get the heroes in both teams
"\n9. Test accuracy vs Number of trees in BaseRF\n"))
except ValueError:
print("Invalid input")
continue
else:
if user_input not in list_of_options:
print("Invalid input")
elif user_input == 1:
draw_heroes_hist(save=True)
elif user_input == 2:
draw_heatmap(save=True)
elif user_input == 3:
web_heatmap()
elif user_input == 4:
which_data = input("Which data to load?\n")
X_matrix, y_matrix, X_train, X_test, y_train, y_test = load_data(which_data)
which_model = input("Which model to load?\n")
model = load_model(which_data + which_model)
graph_name = input("Enter model name\n")
draw_lc_cv(model, graph_name,
10, X_matrix, y_matrix, save=True)
elif user_input == 5:
pass
elif user_input == 6:
pass
elif user_input == 7:
draw_wr_hist(save=True)
elif user_input == 8:
which_data = input("Which data to load?\n")
X_matrix, y_matrix, X_train, X_test, y_train, y_test = load_data(which_data)
which_model = input("XGB or RF?\n")
import time
from src.Models.load_data import *
"""
Measure the time taken for each model to perform 1000 predictions
"""
models = ['Log', 'RF', 'SVC', 'MLP', 'XGB']
results = {}
X_matrix, y_matrix, X_train, X_test, y_train, y_test = load_data('Base')
for model_name in models:
estimator = load_model('Base' + model_name)
start = time.time()
for i in range(1000):
estimator.predict(X_test[i].reshape(1, -1))
runtimes = time.time() - start
results[model_name] = runtimes
print(results)