def test_file(seed, maximum_size, runs, file_extension):
for _ in range(runs):
size = _rd_randint(2, maximum_size)
zeros = _rd_randint(0, size)
mc_to = _MarkovChain.random(size, zeros=zeros, seed=seed)
file_handler, file_path = _tf_mkstemp(suffix=file_extension)
_os_close(file_handler)
# noinspection PyBroadException
try:
mc_to.to_file(file_path)
mc_from = _MarkovChain.from_file(file_path)
exception = False
except Exception:
mc_from = None
exception = True
_os_remove(file_path)
assert exception is False
_npt.assert_allclose(mc_from.p, mc_to.p, rtol=1e-5, atol=1e-8)
def test_dictionary(seed, maximum_size, runs):
for _ in range(runs):
size = _rd_randint(2, maximum_size)
zeros = _rd_randint(0, size)
mc_to = _MarkovChain.random(size, zeros=zeros, seed=seed)
d = mc_to.to_dictionary()
mc_from = _MarkovChain.from_dictionary(d)
_npt.assert_allclose(mc_from.p, mc_to.p, rtol=1e-5, atol=1e-8)
def generate_mc(self, states, window, T):
# set transition
p = np.array([[0.2, 0.8], [0.6, 0.4]])
mc = MarkovChain(p, states)
# set of transions
transitions = mc.walk(int(T / window))
data = []
for i in range(len(transitions)):
for c in self.generate_zipf(float(transitions[i]), window):
data.append(c)
return data
def generate_mc(self, states, window, T, available=False):
# set transition
p = np.array([[0.4, 0.6], [0.75, 0.25]])
mc = MarkovChain(p, states)
# set of transions
transitions = mc.walk(int(T / window))
data = [];
for transition in range(len(transitions)):
for c in self.generate_zipf(float(transitions[transition]), window, transition=transition,
available=available):
data.append(c)
return data
def test_matrix(seed, maximum_size, runs):
_npr.seed(seed)
for _ in range(runs):
size = _rd_randint(2, maximum_size)
m = _npr.randint(101, size=(size, size))
mc1 = _MarkovChain.from_matrix(m)
m = mc1.to_matrix()
mc2 = _MarkovChain.from_matrix(m)
_npt.assert_allclose(mc1.p, mc2.p, rtol=1e-5, atol=1e-8)
def test_identity(size, value):
mc = _MarkovChain.identity(size)
actual = mc.p
expected = _np.asarray(value)
_npt.assert_allclose(actual, expected, rtol=1e-5, atol=1e-8)
def test_graph(seed, maximum_size, runs):
for _ in range(runs):
size = _rd_randint(2, maximum_size)
zeros = _rd_randint(0, size)
mc_to = _MarkovChain.random(size, zeros=zeros, seed=seed)
graph = mc_to.to_graph(False)
mc_from = _MarkovChain.from_graph(graph)
_npt.assert_allclose(mc_from.p, mc_to.p, rtol=1e-5, atol=1e-8)
graph = mc_to.to_graph(True)
mc_from = _MarkovChain.from_graph(graph)
_npt.assert_allclose(mc_from.p, mc_to.p, rtol=1e-5, atol=1e-8)
def test_birth_death(p, q, value):
mc = _MarkovChain.birth_death(p, q)
actual = mc.p
expected = _np.asarray(value)
_npt.assert_allclose(actual, expected, rtol=1e-5, atol=1e-8)
def test_urn_model(n, model, value):
mc = _MarkovChain.urn_model(n, model)
actual = mc.p
expected = _np.asarray(value)
_npt.assert_allclose(actual, expected, rtol=1e-5, atol=1e-8)
def test_fit_walk(fitting_type, possible_states, walk, k, value):
mc = _MarkovChain.fit_walk(fitting_type, possible_states, walk, k)
actual = mc.p
expected = _np.asarray(value)
_npt.assert_allclose(actual, expected, rtol=1e-5, atol=1e-8)
def test_gamblers_ruin(size, w, value):
mc = _MarkovChain.gamblers_ruin(size, w)
actual = mc.p
expected = _np.asarray(value)
_npt.assert_allclose(actual, expected, rtol=1e-5, atol=1e-8)
def test_approximation(size, approximation_type, alpha, sigma, rho, k, value):
mc = _MarkovChain.approximation(size, approximation_type, alpha, sigma,
rho, k)
actual = mc.p
expected = _np.asarray(value)
_npt.assert_allclose(actual, expected, rtol=1e-5, atol=1e-8)
def test_plot_redistributions(seed, maximum_size, maximum_distributions, runs):
rs = _rd_getstate()
_rd_seed(seed)
configs = []
for _ in range(runs):
size = _rd_randint(2, maximum_size)
zeros = _rd_randint(0, size)
configs.append((size, zeros))
_rd_setstate(rs)
mcs = []
plot_types = ['heatmap', 'projection']
for i in range(runs):
size, zeros = configs[i]
mc = _MarkovChain.random(size, zeros=zeros, seed=seed)
if i == 0:
distributions = mc.redistribute(1, output_last=False)
initial_status = None
plot_type = 'projection'
else:
r = _rd_randint(1, maximum_distributions)
distributions = r if _rd_random() < 0.5 else mc.redistribute(r, output_last=False)
initial_status = None if isinstance(distributions, int) or _rd_random() < 0.5 else distributions[0]
plot_type = _rd_choice(plot_types)
configs[i] = (distributions, initial_status, plot_type)
mcs.append(mc)
for i in range(runs):
mc = mcs[i]
distributions, initial_status, plot_type = configs[i]
# noinspection PyBroadException
try:
figure, _ = _plot_redistributions(mc, distributions, initial_status, plot_type)
_mplp.close(figure)
exception = False
except Exception:
exception = True
assert exception is False
def test_steady_states(p, steady_states):
actual = MarkovChain(p).pi
expected = [np.array(steady_state) for steady_state in steady_states]
matches = 0
for a in actual:
for e in expected:
matches += 1 if np.allclose(a, e, rtol=0.0, atol=1e-6) else 0
assert matches == len(expected)
def run(final_result):
print("Calculating frequency of occurance of packets...")
#all unique transition pairs counter
transition_counter = Counter()
for i in range(len(final_result)):
for j in range(len(final_result[i])):
for k in range(len(final_result[i][j]) - 1):
transition_counter[final_result[i][j][k] + "|||" +
final_result[i][j][k + 1]] += 1
#total occurance of the first item in the transition pairs
first_total = Counter()
for i in transition_counter:
first_total[i.split("|||")[0]] += transition_counter[i]
#converting transition_counter to pure frequency in percentage
final_frequency = defaultdict()
for i in transition_counter:
final_frequency[i] = transition_counter[i] / first_total[i.split("|||")
[0]]
#proceed to rule extraction
rule_extraction(final_frequency)
#list of all the nodes in first_total
node_list = []
for i in first_total:
node_list.append(i)
# list of frequencies in matrixs format
p = [[0] * len(node_list) for _ in range(len(node_list))]
for i in node_list:
for j in node_list:
if i + "|||" + j in final_frequency:
p[node_list.index(i)][node_list.index(j)] = final_frequency[
i + "|||" + j]
##node_list rename, save_correspondance to default_dict
symbol_packet_pair = defaultdict()
symbol = 0
print("Converting packet name to symbol...")
for i in range(len(node_list)):
symbol_packet_pair[str(symbol)] = node_list[i]
node_list[i] = str(symbol)
symbol = symbol + 1
print("Writing out symbol_packet_pair...")
with open('extraction_symbol_packet_pair.txt', 'w') as f:
for i in symbol_packet_pair:
f.write(i + " : " + symbol_packet_pair[i] + "\n")
print("Creating markov chain...")
mc = MarkovChain(p, node_list)
print("Plotting MC...")
plt.figure.Figure = plot_graph(mc)[0]
plt.subplots.figure = plot_graph(mc)[1]
plt.savefig("figure.png")
def run(final_result):
transition_counter = Counter()
for i in range(len(final_result)):
for j in range(len(final_result[i])):
for k in range(len(final_result[i][j])-1):
transition_counter[final_result[i][j][k]+"|||"+final_result[i][j][k+1]]+=1
first_total = Counter()
for i in transition_counter:
first_total[i.split("|||")[0]]+=transition_counter[i]
final_frequency = defaultdict()
for i in transition_counter:
final_frequency[i] = transition_counter[i]/first_total[i.split("|||")[0]]
#list of all the nodes
node_list = []
for i in first_total:
node_list.append(i)
# list of frequencies
p = [ [0]*len(node_list) for _ in range(len(node_list))]
for i in node_list:
for j in node_list:
if i+"|||"+j in final_frequency:
p[node_list.index(i)][node_list.index(j)]=final_frequency[i+"|||"+j]
##node_list rename, save_correspondance to default_dict
symbol_packet_pair = defaultdict()
symbol = 0
print("Converting packet name to symbol...")
for i in range(len(node_list)):
symbol_packet_pair[str(symbol)]=node_list[i]
node_list[i]=str(symbol)
symbol = symbol+1
print("Writing out symbol_packet_pair...")
with open ('extraction_symbol_packet_pair.txt', 'w') as f:
for i in symbol_packet_pair:
f.write(i+" : "+symbol_packet_pair[i]+"\n")
mc = MarkovChain(p, node_list)
print(mc)
print(mc.transition_probability('17','22'))
def test_fit_function(possible_states, f, quadrature_type, quadrature_interval,
value):
f = eval('lambda x_index, x_value, y_index, y_value: ' + f)
quadrature_interval = None if quadrature_interval is None else tuple(
quadrature_interval)
mc = _MarkovChain.fit_function(possible_states, f, quadrature_type,
quadrature_interval)
actual = mc.p
expected = _np.asarray(value)
_npt.assert_allclose(actual, expected, rtol=1e-5, atol=1e-8)
def test_plot_walk(seed, maximum_size, maximum_simulations, runs):
rs = _rd_getstate()
_rd_seed(seed)
configs = []
for _ in range(runs):
size = _rd_randint(2, maximum_size)
zeros = _rd_randint(0, size)
configs.append((size, zeros))
_rd_setstate(rs)
mcs = []
plot_types = ['histogram', 'sequence', 'transitions']
for i in range(runs):
size, zeros = configs[i]
mc = _MarkovChain.random(size, zeros=zeros, seed=seed)
r = _rd_randint(2, maximum_simulations)
walk = r if _rd_random() < 0.5 else mc.walk(r, output_indices=True)
initial_state = None if isinstance(walk, int) or _rd_random() < 0.5 else walk[0]
plot_type = _rd_choice(plot_types)
configs[i] = (walk, initial_state, plot_type)
mcs.append(mc)
for i in range(runs):
mc = mcs[i]
walk, initial_state, plot_type = configs[i]
# noinspection PyBroadException
try:
figure, _ = _plot_walk(mc, walk, initial_state, plot_type)
_mplp.close(figure)
exception = False
except Exception:
exception = True
assert exception is False
def test_plot_eigenvalues(seed, maximum_size, runs):
for _ in range(runs):
size = _rd_randint(2, maximum_size)
zeros = _rd_randint(0, size)
mc = _MarkovChain.random(size, zeros=zeros, seed=seed)
# noinspection PyBroadException
try:
figure, _ = _plot_eigenvalues(mc)
_mplp.close(figure)
exception = False
except Exception:
exception = True
assert exception is False
def main():
for file in files:
f = open(file, "r")
lines = f.readlines()
res = [line.split() for line in lines]
cross_tab = transition_matrix(res)
f.close()
# print(file[:-4])
# print(cross_tab)
mc = MarkovChain(cross_tab.values, cross_tab.index._data)
r = open(file.replace("_trans", "_matrix"), "w")
r.write(str(mc.p))
r.close()
# matplotlib.interactive(True)
figure, _ = plot_graph(
mc,
dpi=1000,
path="/home/ja/Dokumenty/Pepper/data_analisys/" +
file.replace("_trans.txt", "_graph.svg"))
plt.close()
def test_plot_graph(seed, maximum_size, runs):
rs = _rd_getstate()
_rd_seed(seed)
configs = []
for _ in range(runs):
size = _rd_randint(2, maximum_size)
zeros = _rd_randint(0, size)
configs.append((size, zeros) + tuple(_rd_random() < 0.5 for _ in range(4)))
_rd_setstate(rs)
for i in range(runs):
size, zeros, nodes_color, nodes_type, edges_color, edges_value = configs[i]
mc = _MarkovChain.random(size, zeros=zeros, seed=seed)
# noinspection PyBroadException
try:
figure, _ = _plot_graph(mc, nodes_color=nodes_color, nodes_type=nodes_type, edges_color=edges_color, edges_value=edges_value, force_standard=True)
_mplp.close(figure)
figure, _ = _plot_graph(mc, nodes_color=nodes_color, nodes_type=nodes_type, edges_color=edges_color, edges_value=edges_value, force_standard=False)
_mplp.close(figure)
exception = False
except Exception:
exception = True
assert exception is False
def test_random(seed, size, zeros, mask, value):
mc = _MarkovChain.random(size, None, zeros, mask, seed)
actual = mc.p
expected = _np.asarray(value)
_npt.assert_allclose(actual, expected, rtol=1e-5, atol=1e-8)
if zeros > 0 and mask is None:
actual = size**2 - _np.count_nonzero(mc.p)
expected = zeros
assert actual == expected
if mask is not None:
indices = ~_np.isnan(_np.asarray(mask))
actual = mc.p[indices]
expected = _np.asarray(value)[indices]
_npt.assert_allclose(actual, expected, rtol=1e-5, atol=1e-8)
from pydtmc import (MarkovChain, plot_graph)
from pytest import (mark)
import matplotlib.pyplot as plt
p = [[0.2, 0.7, 0.0, 0.1], [0.0, 0.6, 0.3, 0.1], [0.0, 0.0, 1.0, 0.0],
[0.5, 0.0, 0.5, 0.0]]
mc = MarkovChain(p, ['s1', 's2', 's3', 's4'])
print(mc)
plt.figure.Figure = plot_graph(mc)[0]
plt.subplots.figure = plot_graph(mc)[1]
#pp.axes._subplots.AxesSubplot = plot_graph(mc)[1]
plt.savefig("figure.png")
print(type(plot_graph(mc)[0]))
print(type(plot_graph(mc)[1]))
#s1/s2/s3/r1/r4/s2/s3/r4/r3/s2/s1/r3/r2/
#counter[s1/r2]: 90%
#counter[s1se1se2/r2]: 80%
def test_periodicity(p, period):
mc = MarkovChain(p)
assert mc.period == period
assert mc.is_aperiodic == (period == 1)