def closure_1_3_4():
k = 4
best_model = None
min_error = np.inf
for i in range(50):
model = Kmedians(k)
model.fit(X)
error = model.error(X)
if error < min_error:
min_error = error
best_model = model
plt.figure()
utils.plot_2dclustering(X, best_model.predict(X))
fname = os.path.join("..", "figs",
"kmedians_outliers_best_model.png")
plt.savefig(fname)
print("\nFigure saved as '%s'" % fname)
plt.plot(k_values,
y,
label="K-Means minimum error plot against k values")
plt.xlabel("k values")
plt.ylabel("Minimum error")
plt.legend()
fname = os.path.join("..", "figs", "q13minerror.pdf")
plt.savefig(fname)
if question == '1.3.4':
X = utils.load_dataset('clusterData2')['X']
model_array = []
error_array = []
for i in range(0, 50):
model = Kmedians(k=4)
model_array.append(model)
model_array[i].fit(X)
model_array[i].predict(X)
error_array.append(model_array[i].error(X))
y = min(error_array)
index = error_array.index(y)
model = model_array[index]
y = model.predict(X)
utils.plot_2dclustering(X, y)
if question == '1.4':
X = utils.load_dataset('clusterData2')['X']
# model_2 = Kmeans(k=4)
# model_2 = Kmedians(k=4)
# model_2.fit(X)
# if model_2.error(X) < error:
# model = model_2
# plot_2dclustering(X, model.predict(X))
#
# fname = os.path.join("..", "figs", "kmeans_3_3.png")
# plt.savefig(fname)
# print("\nFigure saved as '%s'" % fname)
''' Part 2, Part 4 '''
errors = np.ones([10, ]) * -1
for i in range(50):
kVal = random.randint(1, 10)
# model = Kmeans(k=kVal)
model = Kmedians(k=kVal)
model.fit(X)
error = model.error(X)
if errors[kVal - 1] == -1 or error < errors[kVal - 1]:
errors[kVal - 1] = error
print(errors)
plt.plot(range(1, 11), errors)
plt.xlabel('k')
plt.ylabel('Error')
plt.title('Error vs k')
fname = os.path.join("..", "figs", "error_vs_kmedians_3_3.png")
plt.savefig(fname)
print("\nFigure saved as '%s'" % fname)
elif question == '3.4':
# fname3 = "C:\\Users\\wangzhen\\Desktop\\cpsc340\\g5e0b_u7p1b_a2-master\\figs\\q3_3_kmedians.png"
# plt.savefig(fname3)
# print("\nFigure saved as '%s'" % fname3)
X = load_dataset('clusterData2.pkl')['X']
# model=Kmeans(k=1)
# model.fit(X)
# error=model.error(X)
# print(error)
minError = np.zeros(10)
for k in range(10):
minError[k] = np.inf
larg_slope = 0
slope_k = 1
for i in range(50):
model = Kmedians(k=k + 1)
model.fit(X)
error = model.error(X)
if error < minError[k]:
minError[k] = error
print(-minError[k] + minError[k - 1])
if k != 0:
if -minError[k] + minError[k - 1] > larg_slope:
larg_slope = -minError[k] + minError[k - 1]
x1 = [i + 1 for i in range(10)]
y1 = minError
plt.figure(1)
plot1 = plt.plot(x1, y1, 'b', label='minError')
plt.title(
plt.draw()
fname = os.path.join("..", "figs", "kmeans_elbow_method.png")
plt.savefig(fname)
print("\nFigure saved as '%s'" % fname)
#print(errors)
# Using K-medians for different clustering problems
elif module == '4':
X = load_dataset('clusterData2.pkl')['X']
# using elbow method to eyeball the best k
errors = np.zeros(10)
for i in range(10):
min_error = np.inf
for j in range(50):
model = Kmedians(k=i + 1)
model.fit(X)
err = model.error(X)
if err < min_error:
min_error = err
errors[i] = min_error
plt.plot(np.arange(1, 11), errors)
plt.title("k Vs min_error")
plt.xlabel("k")
plt.ylabel("Min error across 50 random initializations")
plt.draw()
fname = os.path.join("..", "figs", "kmedians_elbow_method.png")
plt.savefig(fname)
print("\nFigure saved as '%s'" % fname)
# Trying out scikit's DBSCAN