def main():
# test the performance of the engineered features
# edges list and nodes list file
args = ArgParser.parse_args()
main_path = args.main_path + 'inputs/'
pfilename = args.graph
node_filename = main_path + 'nodes_' + pfilename + '.csv'
edge_filename = main_path + 'edges_' + pfilename + '.csv'
print('Read edge list and node list.')
edges_df = pd.read_csv(edge_filename)
nodes_df = pd.read_csv(node_filename)
# get anchor nodes list
print('Retrieve anchor nodes list.')
anchor_nodes = nodes_df.loc[nodes_df['is_anchor'] == 1]['node'].tolist()
# --- generate features for all anchor nodes of a graph ---
print('Generate features for the anchor nodes.')
n_eng_features = NodeEngFeatures(nodes_df, edges_df)
node_feature_df = n_eng_features.gen_node_features_list(anchor_nodes)
# save anchor nodes features to file
print(node_feature_df.head())
print('Save node features dataframe.')
node_features_filename = main_path + 'features_eth_0.csv'
node_feature_df.to_csv(node_features_filename, index=False)
def runDielec(arg):
"""
Runs UHBD calculation with a set of dielectric constants in dielc_file.
"""
# Parse the command line arguments
dielec = ArgParser.modeClass("dielec",["pdb_file","dielec_file"],
["inpfile","inpfile"],["ionic_strength","pHtitr","out_dir"])
dielec.parseArg(arg)
filename = dielec.required["pdb_file"]
dielec_file = dielec.required["dielec_file"]
output_path = dielec.optional.out_dir
pH_start = dielec.optional.pHtitr[0]
pH_stop = dielec.optional.pHtitr[1]
pH_interval = dielec.optional.pHtitr[2]
salt = dielec.optional.ionic_strength
f = open(dielec_file)
dielectric_constants = [float(d) for d in f.readlines()]
f.close()
base_path = os.path.join(__init__.invocation_path,output_path)
for dielectric in dielectric_constants:
print "dielectric constant: %4.2F" % dielectric
dielec_path = os.path.join(base_path,"D%.1F" % dielectric)
FileOps.makeDir(dielec_path)
dielec_output = os.path.join(dielec_path,"%.1F" % salt)
FileOps.makeDir(dielec_output)
indivRun(filename,dielec_output,pH_start,pH_stop,pH_interval,salt,
dielectric)
def main():
"""
end-to-end classification
"""
binary = True # binary or multi-class classification.
args = ArgParser.parse_args()
input_path = args.input
graph_filename = args.graph
prod_data_dir = args.output
flag = args.flag
clf_opt = args.clf_opt
exp_id = args.exp_id
# data_path = main_path + 'inputs/'
# prod_data_dir = main_path + 'outputs/'
stats_file = prod_data_dir + 'stats_' + flag + '.csv'
edges_filename = input_path + 'edges_' + graph_filename + '.csv'
nodes_filename = input_path + 'nodes_' + graph_filename + '.csv'
features_filename = prod_data_dir + 'features_' + graph_filename + '.csv'
feat_imp_filename = prod_data_dir + 'feature_importance_' + graph_filename + '.csv'
if clf_opt == 'fe':
# ------------------ Feature Engineering ------------------
# read the input file and generating the features and the labels set
print("Node Classification --- Feature Engineering ---")
EF_analysis_selected_nodes(prod_data_dir, graph_filename, edges_filename, nodes_filename,
features_filename, stats_file, feat_imp_filename, 'FE', binary,
rnd_seed, exp_id, extra_analysis=False)
print("--- Node Classification Feature Engineering is done ---")
# ---------------------------------------------------------
elif clf_opt == 'concat':
print("Node classification: Concat. FE &" + flag + " embeddings.")
emb_file = prod_data_dir + 'emb_' + str(flag) + '_' + graph_filename + '.emb'
fe_file = features_filename
nd_clf_fe_emb_combined(emb_file, fe_file, stats_file, flag, binary, exp_id)
else:
# ------------------ RiWalk -------------------------------
print("Node classification: --- RiWalk - " + flag + "---")
# set file names
emb_filename = prod_data_dir + 'emb_' + str(flag) + '_' + graph_filename + '.emb'
RiWalk_analysis_selected_nodes(prod_data_dir, graph_filename, emb_filename, nodes_filename, stats_file,
flag, binary, exp_id, extra_analysis=False)
print("--- Classification RiWalk is done ---")
def runSingle(arg):
"""
Runs UHBD calculation on a single file.
"""
single = ArgParser.modeClass("single",["pdb_file"],["inpfile"],
["dielectric", "ionic_strength","pHtitr","out_dir"])
single.parseArg(arg)
indivRun(single.required["pdb_file"],single.optional.out_dir,
single.optional.pHtitr[0],single.optional.pHtitr[1],
single.optional.pHtitr[2],single.optional.ionic_strength,
single.optional.dielectric)
def runSalts(arg):
"""
Runs UHBD calculation on set of salts in salts_file.
"""
# Parse the command line arguments
salts = ArgParser.modeClass("salts",["pdb_file","salt_file"],
["inpfile","inpfile"],["dielectric","pHtitr","out_dir"])
salts.parseArg(arg)
filename = salts.required["pdb_file"]
salts_file = salts.required["salt_file"]
output_path = salts.optional.out_dir
pH_start = salts.optional.pHtitr[0]
pH_stop = salts.optional.pHtitr[1]
pH_interval = salts.optional.pHtitr[2]
dielectric = salts.optional.dielectric
# Execute the indivRun function for every salt in salts_file
f = open(salts_file)
salts = [float(salt) for salt in f.readlines()]
f.close()
for salt in salts:
print "Ionic strength: %4.2F" % salt
salt_output = os.path.join(output_path,"%.1F" % salt)
# Create output directory
try:
os.mkdir(os.path.join(__init__.invocation_path,salt_output))
except OSError, value:
# Don't stop if we are only overwriting existing directory
if value[0] != 17:
print 'File error.'
print value[0], os.path.join(__init__.invocation_path,
salt_output), value[1]
sys.exit()
# Run UHBD at this specific salt concentration
indivRun(filename,salt_output,pH_start,pH_stop,pH_interval,salt,
dielectric)
def runDielec(arg):
"""
Runs UHBD calculation with a set of dielectric constants in dielc_file.
"""
# Parse the command line arguments
dielec = ArgParser.modeClass("dielec",["pdb_file","dielec_file"],
["inpfile","inpfile"],["ionic_strength","pHtitr","out_dir"])
dielec.parseArg(arg)
filename = dielec.required["pdb_file"]
dielec_file = dielec.required["dielec_file"]
output_path = dielec.optional.out_dir
pH_start = dielec.optional.pHtitr[0]
pH_stop = dielec.optional.pHtitr[1]
pH_interval = dielec.optional.pHtitr[2]
salt = dielec.optional.ionic_strength
f = open(dielec_file)
dielectric_constants = [float(d) for d in f.readlines()]
f.close()
for dielectric in dielectric_constants:
print "dielectric constant: %4.2F" % dielectric
dielec_output = os.path.join(output_path,"%.1F" % dielectric)
# Create output directory
try:
os.mkdir(os.path.join(__init__.invocation_path,dielec_output))
except OSError, value:
# Don't stop if we are only overwriting existing directory
if value[0] != 17:
print 'File error.'
print value[0], os.path.join(__init__.invocation_path,
dielec_output), value[1]
sys.exit()
indivRun(filename,dielec_output,pH_start,pH_stop,pH_interval,salt,
dielectric)
def runSalts(arg):
"""
Runs UHBD calculation on set of salts in salts_file.
"""
# Parse the command line arguments
salts = ArgParser.modeClass("salts",["pdb_file","salt_file"],
["inpfile","inpfile"],["dielectric","pHtitr","out_dir"])
salts.parseArg(arg)
filename = salts.required["pdb_file"]
salts_file = salts.required["salt_file"]
output_path = salts.optional.out_dir
pH_start = salts.optional.pHtitr[0]
pH_stop = salts.optional.pHtitr[1]
pH_interval = salts.optional.pHtitr[2]
dielectric = salts.optional.dielectric
# Execute the indivRun function for every salt in salts_file
f = open(salts_file)
salts = [float(salt) for salt in f.readlines()]
f.close()
base_path = os.path.join(__init__.invocation_path,output_path)
dielec_path = os.path.join(base_path,"D%.1F" % dielectric)
FileOps.makeDir(dielec_path)
for salt in salts:
print "Ionic strength: %4.2F" % salt
salt_output = os.path.join(dielec_path,"%.1F" % salt)
# Create output directory
FileOps.makeDir(salt_output)
# Run UHBD at this specific salt concentration
indivRun(filename,salt_output,pH_start,pH_stop,pH_interval,salt,
dielectric)
def main():
"""
Experiments
"""
# retrieve arguments
args = ArgParser.parse_args()
subgraph_path = args.input
graph_name = args.graph
prod_data_path = args.output
gprep_opt = args.gprep_opt
# input file; directly from XBlock
node_filename = subgraph_path + 'nodes_' + graph_name + '.csv'
edge_filename = subgraph_path + 'edges_' + graph_name + '.csv'
# generated file
nodelist_RiWalk_filename = prod_data_path + 'nodes_' + graph_name + '.nodelist'
edgelist_RiWalk_filename = prod_data_path + 'edges_' + graph_name + '.edgelist'
features_filename = prod_data_path + 'features_' + graph_name + '.csv'
feat_imp_filename = prod_data_path + 'feature_importance_' + graph_name + '.csv'
# elliptic files
elliptic_classes_filename = subgraph_path + '_elliptic_txs_classes.csv'
elliptic_feature_filename = subgraph_path + '_elliptic_txs_features.csv'
nodes = pd.read_csv(node_filename)
edges = pd.read_csv(edge_filename)
# read elliptic dataset files
elliptic_classes = pd.read_csv(elliptic_classes_filename)
elliptic_features = pd.read_csv(elliptic_feature_filename)
tx_features = ["tx_feat_" + str(i) for i in range(2, 95)]
agg_features = ["agg_feat_" + str(i) for i in range(1, 73)]
elliptic_features.columns = ["address", "timestamp"
] + tx_features + agg_features
elliptic_features = pd.merge(elliptic_features,
elliptic_classes,
left_on="address",
right_on="txId",
how='left')
elliptic_features['label'] = elliptic_features['label'].apply(
lambda x: '0' if x == "unknown" else x)
# instantiate a feature engineering object
bc_fe = bfe.BitcoinNodeEngFeatures(nodes, edges)
# graph preparation tasks
if gprep_opt == 'feature':
generate_features_for_all_nodes(bc_fe, elliptic_features,
prod_data_path, graph_name)
elif gprep_opt == 'ri':
# generate edge list for RiWalk
generate_edgelist_for_RiWalk(bc_fe, edgelist_RiWalk_filename)
# generate node list for RiWalk
nodes_all_features_df = pd.read_csv(features_filename)
feature_rank = pd.read_csv(feat_imp_filename)['feature'].tolist()
# selected_features = ['node'] + feature_rank[0:10]
selected_features = ['node'] + feature_rank # all the features
generate_node_list_for_RiWalk(nodes_all_features_df, selected_features,
nodelist_RiWalk_filename)
else:
raise ValueError("Incorrect value for graph preparation option!")
def main():
"""
Experiments
"""
# path setting
binary = True # binary or multi-class classification.ATTENTION: always use BINARY classification
# retrieve arguments from argument parser
args = ArgParser.parse_args()
subgraph_path = args.input
graph_filename = args.graph
prod_data_path = args.output
flag = args.flag
clf_opt = args.clf_opt
exp_id = args.exp_id
stats_file = prod_data_path + 'stats_' + flag + '.csv'
edges_filename = subgraph_path + 'edges_' + graph_filename + '.csv'
nodes_filename = subgraph_path + 'nodes_' + graph_filename + '.csv'
features_filename = prod_data_path + 'features_' + graph_filename + '.csv'
feat_imp_filename = prod_data_path + 'feature_importance_' + graph_filename + '.csv'
if clf_opt == 'fe':
# ------------------ Feature Engineering ------------------
# read the input file and generating the features and the labels set
print("Node Classification --- Feature Engineering ---")
Bitcoin_EF_analysis_selected_nodes(prod_data_path,
graph_filename,
features_filename,
stats_file,
feat_imp_filename,
'FE',
binary,
rnd_seed,
exp_id,
extra_analysis=False)
print("--- Node Classification Feature Engineering is done ---")
# ---------------------------------------------------------
elif clf_opt == 'concat':
# ------------------ FE & Emb concatenating ------------------
print("Node classification: Concat. FE &" + flag + " embeddings.")
emb_file = prod_data_path + 'emb_' + str(
flag) + '_' + graph_filename + '.emb'
fe_file = features_filename
bitcoin_nd_clf_fe_emb_combined(emb_file, fe_file, stats_file, flag,
binary, exp_id)
# ---------------------------------------------------------
else:
# ------------------ RiWalk -------------------------------
print("Node classification: --- RiWalk - " + flag + "---")
# set file names
emb_filename = prod_data_path + 'emb_' + str(
flag) + '_' + graph_filename + '.emb'
nd.RiWalk_analysis_selected_nodes(prod_data_path,
graph_filename,
emb_filename,
nodes_filename,
stats_file,
flag,
binary,
exp_id,
extra_analysis=False)
print("--- Classification RiWalk is done ---")
def main(args=None):
config = configp.start()
if args is None:
args = parser.start()
elif isinstance(args,basestring):
args = parser.start(args)
cedfile = args.ced
stdfile = args.std
plotFields = args.field
multi = args.multi
print_shapes = args.print_shapes
print_global_atts = args.print_global_atts
print_axis = args.print_axis
print_list_synth = args.print_list_synth
terrain = args.terrain
slope = args.slope
meteo = args.meteo
valid = args.valid
prof = args.prof
nearest = args.nearest
noplot = args.no_plot
""" print synhesis availables """
if print_list_synth:
synth_folder=fh.set_working_files(config=config)
out=os.listdir(synth_folder)
print "Synthesis directories available:"
for f in out:
print f
usr_input = raw_input('\nIndicate directory: ')
out=os.listdir(synth_folder+'/'+usr_input)
print "\nSyntheses available:"
out.sort()
for f in out:
if f[-3:]=='cdf': print f
print '\n'
sys.exit()
""" retrieves synthesis and flight instances
from AircraftAnalysis
"""
SYNTH, FLIGHT, TERRAIN = fh.set_working_files(cedfile=cedfile,
stdfile=stdfile,
config=config)
""" print shape of attribute arrays """
if print_shapes:
SYNTH.print_shapes()
if not print_global_atts:
sys.exit()
""" print global attirutes of cedric synthesis """
if print_global_atts:
SYNTH.print_global_atts()
sys.exit()
""" print axis values """
if print_axis:
for ax in print_axis:
if ax.isupper():
ax=ax.lower()
SYNTH.print_axis(ax)
sys.exit()
""" print synthesis time """
# print "Synthesis start time :%s" % SYNTH.start
# print "Synthesis end time :%s\n" % SYNTH.end
""" make synthesis plots """
if plotFields:
for f in plotFields:
P=Plotter.plot_synth(SYNTH,FLIGHT,TERRAIN,
var=f,
wind=args.wind,
panel=args.panel,
slicem = args.slicem,
slicez = args.slicez,
slice = args.slice,
# azimuth = args.azimuth,
# distance = args.distance,
zoomIn=args.zoomin,
mask = args.mask,
config=config)
""" make terrain plots """
if terrain or slope:
# P[0] might produce error if P is not a list,
# check in ploth_synth.cross_section
Plotter.plot_terrain(P[0],
terrain=terrain,
slope=slope,
terrain_file=config['filepath_dtm'])
""" make flight level meteo plot """
if meteo:
Plotter.plot_flight_meteo(SYNTH,FLIGHT)
""" compare synth and flight level """
if valid:
out = Plotter.compare_synth_flight(SYNTH,FLIGHT,
level=valid,
zoomin=args.zoomin,
noplot=noplot)
# if 'wind_profiler' in config:
# case=int(cedfile[1:3])
# Plotter.compare_with_windprof(SYNTH,
# location=config['wind_profiler'],
# case=case)
return out
""" make profile from synthesis """
if prof:
if nearest is None:
markers=['o','s','D','*']
out = Plotter.make_synth_profile(SYNTH,coords=prof,
markers=markers,
noplot=noplot)
else:
' (4.5km,n=12) or (7.0km,n=30) seem good choices '
out = Plotter.make_synth_profile_withnearest(SYNTH,
target_latlon=prof,
max_dist=nearest[0][0], # [km]
n_neigh =nearest[0][1])
try:
' if P exists '
for i,p in enumerate(prof):
lat,lon = p
P.haxis.scatter(lon,lat,
s=config['sounding_size'],
c=config['sounding_color'],
marker=config['sounding_marker'],
lw=3)
except UnboundLocalError:
' if P does not exist just pass '
pass
return out
# if turbulence:
# Plotter.print_covariance(SYNTH,FLIGHT)
# Plotter.print_correlation(SYNTH,FLIGHT)
# Plotter.plot_wind_comp_var(SYNTH,FLIGHT)
# Plotter.plot_tke(SYNTH,FLIGHT)
# Plotter.plot_vertical_heat_flux(SYNTH,FLIGHT)
# Plotter.plot_vertical_momentum_flux(SYNTH,FLIGHT,config['filepath_dtm'])
# Plotter.plot_turbulence_spectra(SYNTH,FLIGHT)
if multi:
plt.close('all')
return P
else:
''' use this one with ipython '''
plt.show(block=False)
''' use this one with the shell '''
# if turbulence:
# Plotter.print_covariance(SYNTH,FLIGHT)
# Plotter.print_correlation(SYNTH,FLIGHT)
# Plotter.plot_wind_comp_var(SYNTH,FLIGHT)
# Plotter.plot_tke(SYNTH,FLIGHT)
# Plotter.plot_vertical_heat_flux(SYNTH,FLIGHT)
# Plotter.plot_vertical_momentum_flux(SYNTH,FLIGHT,config['filepath_dtm'])
# Plotter.plot_turbulence_spectra(SYNTH,FLIGHT)
if multi:
plt.close('all')
return P
else:
''' use this one with ipython '''
plt.show(block=False)
''' use this one with the shell '''
# plt.show()
"""call main function """
if __name__ == "__main__":
args = parser.start()
p = main(args)
import ArgParser import sys Parser = ArgParser.ArgParser(sys.argv) print(Parser.ParsedArgs)
def main():
"""
instantiate a node2vec object
"""
print("Node2Vec main method.")
start_time = time.time()
args = ArgParser.parse_args()
main_path = args.main_path
graph_name = args.graph
iter_num = args.iter
num_walks = args.num_walks
dim = args.dimensions
walk_length = args.walk_length
workers = args.workers
window_size = args.window_size
p = args.p
q = args.q
data_path = main_path + '/inputs/'
edge_list_file_name = data_path + graph_name + '.edgelist'
node_list_file_name = data_path + 'nodes_' + graph_name + '.csv'
output_path = main_path + '/outputs/'
stats_file = output_path + 'stats.csv'
nx_g = nx.from_pandas_edgelist(pd.read_csv(edge_list_file_name),
source='source',
target='target',
create_using=nx.DiGraph())
print("Graph info:", nx.info(nx_g))
# for edge in nx_g.edges():
# nx_g[edge[0]][edge[1]]['weight'] = 1
# nx_g = nx_g.to_undirected()
print("\tInstantiate a node2vec object.")
node2vec = Node2Vec(nx_g,
dimensions=dim,
walk_length=walk_length,
num_walks=num_walks,
workers=workers,
p=p,
q=q)
print("\tFit node2vec.")
model = node2vec.fit(window=window_size,
sg=1,
hs=0,
min_count=1,
iter=iter_num)
# read node list
print("\tExtract embeddings and labels for the anchor nodes.")
nodes_df = pd.read_csv(node_list_file_name)
# binary classification anchor nodes
anchor_nodes_df = nodes_df.loc[nodes_df['is_anchor'] == 1, ['node', 'isp']]
node_list = [str(node_id) for node_id in anchor_nodes_df['node'].tolist()]
embeddings = [model.wv.get_vector(node) for node in node_list]
labels = anchor_nodes_df['isp'].tolist()
# classification
print("\tApply classification.")
rnd_seed = 42
binary = True
X_train, X_test, y_train, y_test = NodeClassification.train_test_split(
embeddings, labels, rnd_seed)
NodeClassification.rf_lr_classification(X_train,
X_test,
y_train,
y_test,
stats_file,
'n2v',
binary=binary,
print_report=True)
print("Total elapsed time:", str(time.time() - start_time))
