json.dump(rich, fout, ensure_ascii=False)
# coalitions
coalitions = []
majority = 101
def _coalition_probability(parties, party_codes, runs, majority):
t = parties.loc[:, party_codes]
# note: all members of coalition has to have >0 seats
return ((t.sum(axis=1) >= majority) * np.sign(
(parties.loc[:, party_codes]).product(axis=1))).sum() / runs
# single parties
for row in stats.iterrows():
it = {
'party_code':
row[1]['party_code'],
'seats':
row[1]['seats'],
'majority_probability':
_coalition_probability(parties, [row[1]['party_code']], runs,
majority),
'party_codes': [row[1]['party_code']]
}
if it['majority_probability'] > 0:
coalitions.append(it)
# two parties
for p in ps:
def tumbling_statistics_per_node(path, tracefile, window_size=10):
# Compute a dictionary containing all the statistics from each node of the dataset
# Read the rank of each node
nodes = pd.read_csv(path + 'addr-' + tracefile + '.cap',
sep=';|seq=| hop|time = |ms',
na_filter=True,
usecols=[1, 3, 5],
header=None,
skiprows=799,
names=['node_id', 'seq', 'rtt'],
engine='python').dropna().drop_duplicates()
nodes = nodes.sort_values(by=['node_id', 'seq'],
ascending=True,
na_position='first')
nodes = nodes[nodes['rtt'] >= 1] # Removes values with RTT < 1ms
d_nodes = {} # <node_id, DataFrame containing seq and rtt columns>
for n in nodes.index:
if nodes['node_id'][n] in d_nodes:
d_nodes[nodes['node_id'][n]] = d_nodes[nodes['node_id'][n]].append(
pd.DataFrame({
'seq': [int(nodes['seq'][n])],
nodes['node_id'][n]: [nodes['rtt'][n]]
}))
else:
d_nodes[nodes['node_id'][n]] = pd.DataFrame({
'seq': [int(nodes['seq'][n])],
nodes['node_id'][n]: [nodes['rtt'][n]]
})
# Generate a dataframe containing all nodes
nodes = pd.DataFrame([seq for seq in range(1, 1001)],
columns=['seq']).set_index('seq')
for node in d_nodes.keys():
nodes = nodes.join(d_nodes[node].set_index('seq'))
nodes = nodes[~nodes.index.duplicated(keep='first')]
# Calculate all the statistics
statistics = {} # <node_id, statistics of the node>
for node in nodes:
stats = nodes[node].groupby(nodes[node].index //
window_size).count().to_frame()
stats = stats.rename(index=str, columns={node: "packet_loss"})
stats["packet_loss"] = pd.to_numeric(stats["packet_loss"],
downcast='float')
for index, row in stats.iterrows():
stats.at[index, 'packet_loss'] = row['packet_loss'] / window_size
stats = stats.assign(
mean=nodes[node].groupby(nodes[node].index //
window_size).mean().values)
stats = stats.assign(min=nodes[node].groupby(nodes[node].index //
window_size).min().values)
stats = stats.assign(max=nodes[node].groupby(nodes[node].index //
window_size).max().values)
stats = stats.assign(std=nodes[node].groupby(nodes[node].index //
window_size).std().values)
stats = stats.assign(var=nodes[node].groupby(nodes[node].index //
window_size).var().values)
statistics[node] = stats
return statistics