def get_frac_nodes_in_gcc_from_dates(start, end):
g_slice = graphgen.get_graph_slice(start, end)
nodes_in_gcc = len(nx.weakly_connected_components(g_slice)[0])
return float(nodes_in_gcc) / g_slice.number_of_nodes()
def get_out_degrees_from_dates(start, end):
g_slice = graphgen.get_graph_slice(start, end)
return g_slice.out_degree()
def get_num_nodes_edges_from_dates(start, end):
g_slice = graphgen.get_graph_slice(start, end)
return g_slice.number_of_nodes(), g_slice.number_of_edges()
def analyze_public_trading():
"""
Analyzes the snippet of the graph around the time when bitcoin opened
for public trading with USD.
"""
g = graphgen.get_graph_slice(_START_PUBLIC_TRADING, _END_PUBLIC_TRADING)
import networkx as nx
import graphgen
import graphtools
import tags_over_time
from networkx.algorithms import *
_HMS = 1000000
slices = graphgen.generate_time_slices(slice_time=5, num_intervals=50, num_in_slice=10)
i = 0
deg_connectivity = []
dates = []
avg_clust = []
lccs = []
largest_scc = []
for start, end in slices:
g = graphgen.get_graph_slice(start * _HMS, end * _HMS)
if len(g) == 0:
continue
# stamp = str(start) + '_' + str(end)
#tags_over_time.user_transaction_frequency(g, stamp)
# tags_over_time.user_transaction_amount(g, stamp)
# tags_over_time.user_buy_frequency(g, stamp)
# tags_over_time.user_sell_frequency(g, stamp)
# d = assortativity.average_degree_connectivity(g)
# utils.save_node_map(d, stamp)
deg_connectivity.append(assortativity.average_degree_connectivity(g))
undir_g = nx.Graph(g.copy())
avg_clust.append(nx.average_clustering(undir_g))
lccs.append(len(graphtools.get_lcc_from_graph(g)))
largest_scc.append(len(graphtools.get_sccs_from_graph(g)[0]))
def analyze_silk_road():
"""
Analyzes the snippet of the bitcoin network around the time that
the silk road launched.
"""
g = graphgen.get_graph_slice(_START_SILK_ROAD, _END_SILK_ROAD)
def analyze_satoshi_dice():
"""
Analyzes the snippet of the bitcoin network around the time that
SatoshiDice was announced. There is a hard date for its launch.
"""
g_before = graphgen.get_graph_slice(_START_SATOSHI_DICE, _LAUNCH_SATOSHI_DICE)
g_after = graphgen.get_graph_slice(_START_SATOSHI_DICE, _END_SATOSHI_DICE)
before_lcc = graphtools.get_lcc_from_graph(g_before)
after_lcc = graphtools.get_lcc_from_graph(g_after)
print len(before_lcc), len(after_lcc)
'''
before_in = g_before.in_degree()
before_in_values = sorted(set(before_in.values()))
before_in_hist = [before_in.values().count(x)/float(len(g_before.nodes())) for x in before_in_values]
after_in = g_after.in_degree()
after_in_values = sorted(set(after_in.values()))
after_in_hist = [after_in.values().count(x)/float(len(g_after.nodes())) for x in after_in_values]
plt.loglog(before_in_values, before_in_hist, 'bo', label='In-degree before SatoshiDICE announced')
plt.loglog(after_in_values, after_in_hist, 'ro', label='In-degree after SatoshiDICE announced')
plt.legend(loc=1)
plt.xlabel('Degree')
plt.ylabel('Fraction of nodes')
plt.title('In-degree 1 week before and after SatoshiDICE announced')
plt.show()
before_out = g_before.out_degree()
before_out_values = sorted(set(before_out.values()))
before_out_hist = [before_out.values().count(x)/float(len(g_before.nodes())) for x in before_out_values]
after_out = g_after.out_degree()
after_out_values = sorted(set(after_out.values()))
after_out_hist = [after_out.values().count(x)/float(len(g_after.nodes())) for x in after_out_values]
plt.loglog(before_out_values, before_out_hist, 'bo', label='Out-degree before SatoshiDICE announced')
plt.loglog(after_out_values, after_out_hist, 'ro', label='Out-degree after SatoshiDICE announced')
plt.legend(loc=1)
plt.xlabel('Degree')
plt.ylabel('Fraction of nodes')
plt.title('Out-degree 1 week before and after SatoshiDICE announced')
plt.show()
'''
#Can't use k-core with MultiDiGraph
'''
g_before.remove_edges_from(g_before.selfloop_edges())
g_after.remove_edges_from(g_after.selfloop_edges())
print len(graphtools.get_k_core_from_graph(g_before).nodes())
print len(graphtools.get_k_core_from_graph(g_after).nodes())
'''
#before_node_max_in = graphtools.get_node_max_in_degree(g_before)
#after_node_max_in = graphtools.get_node_max_in_degree(g_after)
#print before_node_max_in, after_node_max_in
#print sorted(g_before.in_degree().iteritems(), key=operator.itemgetter(1))
#print sorted(g_after.in_degree().iteritems(), key=operator.itemgetter(1))
#print len(g_after.edges('25', data=True))
#print nx.get_edge_attributes(g_after, 'transaction_key')
#print g_after.in_degree(after_node_max_in)
#new_nodes = set(g_after.in_degree())-set(g_before.in_degree())
#print len(new_nodes)
#new_nodes_degrees = {k: g_after.in_degree()[k] for k in new_nodes}
#print sorted(new_nodes_degrees.iteritems(), key=operator.itemgetter(1))
time_period = graphgen._days[graphgen._days.index(_START_SATOSHI_DICE/graphgen._HMS):graphgen._days.index(_END_SATOSHI_DICE/graphgen._HMS)]
transactions_per_day = []
for day in time_period:
daystart = graphtools.string_to_datetime(str(day) + '000000')
dayend = graphtools.string_to_datetime(str(day) + '235959')
transactions = []
for n, nbrs in g_after.adjacency_iter():
for nbr, edict in nbrs.items():
for eattr in edict.values():
timestamp = graphtools.string_to_datetime(eattr['date'])
if (timestamp >= daystart and timestamp <= dayend):
transactions.append(1)
transactions_per_day.append(len(transactions))
print time_period
print transactions_per_day
time_period = [graphtools.string_to_datetime(str(time)) for time in time_period]
dates = matplotlib.dates.date2num(time_period)
plt.plot(time_period, transactions_per_day, 'b-')
plt.xlabel('Date')
plt.ylabel('Number of transactions')
plt.title('Number of transactions 1 week before and after SatoshiDICE announced')
plt.show()
def get_frac_nodes_in_gcc_from_dates(start, end):
g_slice = graphgen.get_graph_slice(start, end)
nodes_in_gcc = len(nx.weakly_connected_components(g_slice)[0])
return float(nodes_in_gcc) / g_slice.number_of_nodes()
def get_effective_diameter_from_dates(start, end):
g_slice = graphgen.get_graph_slice(start, end)
return effective_diameter(g_slice)
def get_out_degrees_from_dates(start, end):
g_slice = graphgen.get_graph_slice(start, end)
return g_slice.out_degree()
def get_num_nodes_edges_from_dates(start, end):
g_slice = graphgen.get_graph_slice(start, end)
return g_slice.number_of_nodes(), g_slice.number_of_edges()
def get_avg_out_degree_from_dates(start, end):
g_slice = graphgen.get_graph_slice(start, end)
return get_avg_out_degree_from_graph(g_slice)
def get_sccs_from_dates(start, end):
g_slice = graphgen.get_graph_slice(start, end)
return nx.strongly_connected_components(graph)
def get_wccs_from_dates(start, end):
g_slice = graphgen.get_graph_slice(start, end)
return nx.weakly_connected_components(g_slice)
def get_effective_diameter_from_dates(start, end):
g_slice = graphgen.get_graph_slice(start, end)
return effective_diameter(g_slice)
import graphtools
import tags_over_time
from networkx.algorithms import *
_HMS = 1000000
slices = graphgen.generate_time_slices(slice_time=5,
num_intervals=50,
num_in_slice=10)
i = 0
deg_connectivity = []
dates = []
avg_clust = []
lccs = []
largest_scc = []
for start, end in slices:
g = graphgen.get_graph_slice(start * _HMS, end * _HMS)
if len(g) == 0:
continue
# stamp = str(start) + '_' + str(end)
#tags_over_time.user_transaction_frequency(g, stamp)
# tags_over_time.user_transaction_amount(g, stamp)
# tags_over_time.user_buy_frequency(g, stamp)
# tags_over_time.user_sell_frequency(g, stamp)
# d = assortativity.average_degree_connectivity(g)
# utils.save_node_map(d, stamp)
deg_connectivity.append(assortativity.average_degree_connectivity(g))
undir_g = nx.Graph(g.copy())
avg_clust.append(nx.average_clustering(undir_g))
lccs.append(len(graphtools.get_lcc_from_graph(g)))
largest_scc.append(len(graphtools.get_sccs_from_graph(g)[0]))
def _node_and_edges(start, end):
g = graphgen.get_graph_slice(start * 1000000, end * 1000000)
if len(g) == 0:
return 0, 0
return g.number_of_nodes(), g.number_of_edges()
def get_wccs_from_dates(start, end):
g_slice = graphgen.get_graph_slice(start, end)
return nx.weakly_connected_components(g_slice)
def analyze_public_trading():
"""
Analyzes the snippet of the graph around the time when bitcoin opened
for public trading with USD.
"""
g = graphgen.get_graph_slice(_START_PUBLIC_TRADING, _END_PUBLIC_TRADING)
def get_sccs_from_dates(start, end):
g_slice = graphgen.get_graph_slice(start, end)
return nx.strongly_connected_components(graph)
def analyze_silk_road():
"""
Analyzes the snippet of the bitcoin network around the time that
the silk road launched.
"""
g = graphgen.get_graph_slice(_START_SILK_ROAD, _END_SILK_ROAD)
def get_avg_out_degree_from_dates(start, end):
g_slice = graphgen.get_graph_slice(start, end)
return get_avg_out_degree_from_graph(g_slice)
def analyze_satoshi_dice():
"""
Analyzes the snippet of the bitcoin network around the time that
SatoshiDice was announced. There is a hard date for its launch.
"""
g_before = graphgen.get_graph_slice(_START_SATOSHI_DICE,
_LAUNCH_SATOSHI_DICE)
g_after = graphgen.get_graph_slice(_START_SATOSHI_DICE, _END_SATOSHI_DICE)
before_lcc = graphtools.get_lcc_from_graph(g_before)
after_lcc = graphtools.get_lcc_from_graph(g_after)
print len(before_lcc), len(after_lcc)
'''
before_in = g_before.in_degree()
before_in_values = sorted(set(before_in.values()))
before_in_hist = [before_in.values().count(x)/float(len(g_before.nodes())) for x in before_in_values]
after_in = g_after.in_degree()
after_in_values = sorted(set(after_in.values()))
after_in_hist = [after_in.values().count(x)/float(len(g_after.nodes())) for x in after_in_values]
plt.loglog(before_in_values, before_in_hist, 'bo', label='In-degree before SatoshiDICE announced')
plt.loglog(after_in_values, after_in_hist, 'ro', label='In-degree after SatoshiDICE announced')
plt.legend(loc=1)
plt.xlabel('Degree')
plt.ylabel('Fraction of nodes')
plt.title('In-degree 1 week before and after SatoshiDICE announced')
plt.show()
before_out = g_before.out_degree()
before_out_values = sorted(set(before_out.values()))
before_out_hist = [before_out.values().count(x)/float(len(g_before.nodes())) for x in before_out_values]
after_out = g_after.out_degree()
after_out_values = sorted(set(after_out.values()))
after_out_hist = [after_out.values().count(x)/float(len(g_after.nodes())) for x in after_out_values]
plt.loglog(before_out_values, before_out_hist, 'bo', label='Out-degree before SatoshiDICE announced')
plt.loglog(after_out_values, after_out_hist, 'ro', label='Out-degree after SatoshiDICE announced')
plt.legend(loc=1)
plt.xlabel('Degree')
plt.ylabel('Fraction of nodes')
plt.title('Out-degree 1 week before and after SatoshiDICE announced')
plt.show()
'''
#Can't use k-core with MultiDiGraph
'''
g_before.remove_edges_from(g_before.selfloop_edges())
g_after.remove_edges_from(g_after.selfloop_edges())
print len(graphtools.get_k_core_from_graph(g_before).nodes())
print len(graphtools.get_k_core_from_graph(g_after).nodes())
'''
#before_node_max_in = graphtools.get_node_max_in_degree(g_before)
#after_node_max_in = graphtools.get_node_max_in_degree(g_after)
#print before_node_max_in, after_node_max_in
#print sorted(g_before.in_degree().iteritems(), key=operator.itemgetter(1))
#print sorted(g_after.in_degree().iteritems(), key=operator.itemgetter(1))
#print len(g_after.edges('25', data=True))
#print nx.get_edge_attributes(g_after, 'transaction_key')
#print g_after.in_degree(after_node_max_in)
#new_nodes = set(g_after.in_degree())-set(g_before.in_degree())
#print len(new_nodes)
#new_nodes_degrees = {k: g_after.in_degree()[k] for k in new_nodes}
#print sorted(new_nodes_degrees.iteritems(), key=operator.itemgetter(1))
time_period = graphgen._days[graphgen._days.index(_START_SATOSHI_DICE /
graphgen._HMS):graphgen.
_days.index(_END_SATOSHI_DICE /
graphgen._HMS)]
transactions_per_day = []
for day in time_period:
daystart = graphtools.string_to_datetime(str(day) + '000000')
dayend = graphtools.string_to_datetime(str(day) + '235959')
transactions = []
for n, nbrs in g_after.adjacency_iter():
for nbr, edict in nbrs.items():
for eattr in edict.values():
timestamp = graphtools.string_to_datetime(eattr['date'])
if (timestamp >= daystart and timestamp <= dayend):
transactions.append(1)
transactions_per_day.append(len(transactions))
print time_period
print transactions_per_day
time_period = [
graphtools.string_to_datetime(str(time)) for time in time_period
]
dates = matplotlib.dates.date2num(time_period)
plt.plot(time_period, transactions_per_day, 'b-')
plt.xlabel('Date')
plt.ylabel('Number of transactions')
plt.title(
'Number of transactions 1 week before and after SatoshiDICE announced')
plt.show()
def _node_and_edges(start, end):
g = graphgen.get_graph_slice(start * 1000000, end * 1000000)
if len(g) == 0:
return 0, 0
return g.number_of_nodes(), g.number_of_edges()
