def __init__(self, raw_data : pd.DataFrame, n_clusters : int) -> None:
"""Initialize the object.
Arguments:
raw_data -- pandas DataFrame with timestamp as index.
n_clusters -- number of clusters to be used in clustering.
"""
self.graph = StateGraph(n_clusters)
self.data = self.graph.fit_transform(raw_data)
self.norm_data = pd.DataFrame(
data=self.graph.normalisation.transform(raw_data),
index=raw_data.index,
columns=raw_data.columns)
self.labels = np.array([self.data["label"].to_numpy()]).T
class Visualization():
"""
Maintains information of a single dataset
and provides methods for its visualization.
"""
def __init__(self, raw_data: pd.DataFrame, n_clusters: int) -> None:
"""Initialize the object.
Arguments:
raw_data -- pandas DataFrame with timestamp as index.
n_clusters -- number of clusters to be used in clustering.
"""
self.graph = StateGraph(n_clusters)
data = self.graph.fit_transform(raw_data)
self.norm_data = pd.DataFrame(
data=self.graph.normalisation.transform(raw_data),
index=raw_data.index,
columns=raw_data.columns)
self.labels = np.array([data["label"].to_numpy()]).T
def get_parallel_plot(self) -> go.Figure:
"""Returns a Figure containing the parallel plot of
clusters' centroids.
"""
fig = px.parallel_coordinates(self.graph.centroids,
color=self.graph.centroids.index,
dimensions=self.graph.centroids.columns)
return fig
def get_PCA(self) -> go.Figure:
"""Returns a Figure containing a PCA plot of clustered data."""
pca = PCA(n_components=2)
pca_result = pca.fit_transform(self.norm_data)
plot_data = pd.DataFrame(np.concatenate((pca_result, self.labels),
axis=1),
columns=["pca_x", "pca_y", "cluster"])
return px.scatter(plot_data, x="pca_x", y="pca_y", color="cluster")
def get_TSNE(self, perplexity: int = 100, n_iter: int = 3000) -> go.Figure:
"""Returns a Figure containing a TSNE plot of clustered data.
Function arguments correspond to standard TSNE arguments.
"""
tsne = TSNE(n_components=2,
verbose=1,
perplexity=perplexity,
n_iter=n_iter)
tsne_result = tsne.fit_transform(self.norm_data)
plot_data = pd.DataFrame(np.concatenate((tsne_result, self.labels),
axis=1),
columns=["tsne_x", "tsne_y", "cluster"])
return px.scatter(plot_data, x="tsne_x", y="tsne_y", color="cluster")
def get_all_graphs_infos(cls):
for (_, _, filenames) in os.walk(FOLDER_GRAPH):
for filename in filenames:
if '.gz' in filename:
graph_name = filename.replace('.gz',
'') # Remove .gz extension.
domain_name = filename.split('-', 1)[0]
options_list = [
'--domain', domain_name, '--graph', graph_name
]
exp_params = ExpParams.get_exp_params(options_list)
g = StateGraph.load(exp_params)
g.print_stats()
break
def create_back_range(cls, exp_params):
g = StateGraph.load(exp_params)
g.compute_bfs()
# prob_keep_range = [1.0, 0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2, 0.1, 0.0]
prob_keep_range = [x / 100.0 for x in EXP_BACK_RANGE]
prob_keep_range_inc = []
currently_kept = 1.0
for p in prob_keep_range:
must_keep_inc = p / currently_kept
prob_keep_range_inc.append(must_keep_inc)
currently_kept = p
# print 'Going to trim back edges by these percentages:'
# print prob_keep_range_inc
for i in range(len(prob_keep_range)): # pylint: disable=consider-using-enumerate
filename_suffix = '-back-%d' % (prob_keep_range[i] * 100)
g.trim_back_edges(prob_keep_range_inc[i])
g.save(exp_params, filename_suffix)
print ''
class Visualization():
"""
Maintains information of a single dataset
and provides methods for its visualization.
"""
def __init__(self, raw_data : pd.DataFrame, n_clusters : int) -> None:
"""Initialize the object.
Arguments:
raw_data -- pandas DataFrame with timestamp as index.
n_clusters -- number of clusters to be used in clustering.
"""
self.graph = StateGraph(n_clusters)
self.data = self.graph.fit_transform(raw_data)
self.norm_data = pd.DataFrame(
data=self.graph.normalisation.transform(raw_data),
index=raw_data.index,
columns=raw_data.columns)
self.labels = np.array([self.data["label"].to_numpy()]).T
def get_parallel_plot(self) -> go.Figure:
"""Returns a Figure containing the parallel plot of
clusters' centroids.
"""
fig = px.parallel_coordinates(
self.graph.centroids,
color=self.graph.centroids.index,
dimensions=self.graph.centroids.columns
)
return fig
def get_PCA(self) -> go.Figure:
"""Returns a Figure containing a PCA plot of clustered data."""
pca = PCA(n_components=2)
pca_result = pca.fit_transform(self.norm_data)
plot_data = pd.DataFrame(
np.concatenate((pca_result,self.labels),axis=1),
columns=["pca_x","pca_y","cluster"])
return px.scatter(plot_data,x="pca_x",y="pca_y",color="cluster")
def get_TSNE(self, perplexity : int = 100, n_iter : int = 3000) -> go.Figure:
"""Returns a Figure containing a TSNE plot of clustered data.
Function arguments correspond to standard TSNE arguments.
"""
tsne = TSNE(n_components=2,
verbose=1,
perplexity=perplexity,
n_iter=n_iter)
tsne_result = tsne.fit_transform(self.norm_data)
plot_data = pd.DataFrame(
np.concatenate((tsne_result,self.labels),axis=1),
columns=["tsne_x","tsne_y","cluster"])
return px.scatter(plot_data,x="tsne_x",y="tsne_y",color="cluster")
def get_histograms(self,cluster : int) -> go.Figure:
"""Returns a figure containing a histogram for each sensor
for the data in a given cluster (cluster index starts at 0).
"""
if cluster not in self.labels.flatten():
raise ValueError(f"cluster must be between 0 and {np.amax(self.labels)}")
filtered = self.data[self.data["label"] == cluster]
cols = filtered.columns.values[:-1]
fig = make_subplots(rows=len(cols)//2,cols=2,subplot_titles=cols)
for index, sensor in enumerate(cols):
x = filtered[sensor].values
fig.append_trace(
go.Histogram(x=x, name=sensor),
row=index//2+1,
col=index%2+1
)
fig.update_xaxes(title_text=sensor, row=index//2+1, col=index%2+1)
fig.update_yaxes(title_text="Count", row=index//2+1, col=index%2+1)
fig.update_layout(autosize=False,width=1500,height=200*len(cols),
showlegend=False)
return fig
from sys import stdin, argv
from os import system, name
from state_graph import StateGraph
from maze import Maze
def clear_screen():
if name == 'posix':
system('clear')
else:
system('cls')
try:
with open(argv[1], 'r') as file:
graph = StateGraph(file.read())
except:
print("Falha ao ler o arquivo, verifique o formato passado")
exit()
clear_screen()
print("Escolha um dos algoritmos disponíveis:")
print("1 - A*")
print("2 - Busca em Aprofundamento Iterativo")
try:
while True:
inp = input()
if inp == '1':
path, cost = graph.astar()
# print prob_keep_range_inc
for i in range(len(prob_keep_range)): # pylint: disable=consider-using-enumerate
filename_suffix = '-back-%d' % (prob_keep_range[i] * 100)
g.trim_back_edges(prob_keep_range_inc[i])
g.save(exp_params, filename_suffix)
print ''
@classmethod
def get_all_graphs_infos(cls):
for (_, _, filenames) in os.walk(FOLDER_GRAPH):
for filename in filenames:
if '.gz' in filename:
graph_name = filename.replace('.gz',
'') # Remove .gz extension.
domain_name = filename.split('-', 1)[0]
options_list = [
'--domain', domain_name, '--graph', graph_name
]
exp_params = ExpParams.get_exp_params(options_list)
g = StateGraph.load(exp_params)
g.print_stats()
break
if __name__ == '__main__':
exp_params = ExpParams.get_exp_params_from_command_line_args()
StateGraph.print_graph_info(exp_params)
# ManipGraph.create_back_range(exp_params)
# ManipGraph.get_all_graphs_infos()
s = item[1]
k_score = self.compute_k_neighbours_fittness(
s, self.recent_restores_size)
if s.fitness_score > highest_score:
fittest_node = s
highest_score = s.fitness_score
print s.uid + " k_neighbour_fitness: " + str(s)
return fittest_node.uid
# Some tests
if __name__ == '__main__':
# Initial case -> empty
graph = StateGraph()
graph.add_node('1')
node = graph.retrieve('1')
node.fitness_score = 100
graph.dump()
strategy = CircularRestoreStrategy(graph, 3)
print strategy.get_fittest_state() # expected state 1
print strategy
print strategy.get_fittest_state() # expected state 1
print strategy
# Building up, graph has less states than size of queue
graph.add_node('2')
node = graph.retrieve('2')
node.fitness_score = 90