masterthesis/codes/node-ranking/rank_with_churn_edges.py
2022-04-05 10:23:45 +02:00

143 lines
5.0 KiB
Python

#!/usr/bin/env python3
import numpy as np
from collections import defaultdict
from datetime import datetime
from functools import reduce
from node_ranking import (
page_rank,
sensor_rank,
find_rank,
parse_csv,
csv_loader,
build_graph,
Node,
RankedNode,
)
def load_data(path):
data = defaultdict(list)
with open(path, 'r') as f:
for line in f.readlines():
when = datetime.strptime(line.split(',')[0]+'00', '%Y-%m-%d %H:%M:%S%z')
data[when].append(
parse_csv(line, source_ip_index=1, source_port_index=2, dest_ip_index=3, dest_port_index=4)
)
return data
churned_peers = 50
iterations = 5
def main():
initial_rank = 0.5
data = load_data('./part-dist-edges.csv')
print(f'loaded data. {len(data.keys())} buckets')
for bucket, edges in data.items():
edges = list(edges)
print(f'bucket: {bucket}')
print(f'edges: {len(edges)}')
known = Node('34.204.196.211', 9796)
known_r = RankedNode(known, initial_rank)
g = build_graph(edges, initial_rank=initial_rank)
cnt = 0
destinations = []
for node in g:
destinations.append(node)
cnt += 1
if cnt >= churned_peers:
break
for node in destinations:
g.add_edge(known_r, node)
count_map = {}
for node in g:
count_map[node] = len(list(g.successors(node)))
sum_out = 0
for v in count_map.values():
sum_out += v
min_out = min(count_map.items(), key=lambda kv: kv[1])
max_out = max(count_map.items(), key=lambda kv: kv[1])
nodes = len(g)
initial_rank = 0.5
with open(f'./churn_rank/{churned_peers}_edges_{bucket.timestamp()}.txt', 'w') as out:
table = []
pr_data = []
sr_data = []
for iteration in range(iterations):
g = page_rank(g)
avg_pr = reduce(lambda acc, n: acc + n.rank, g, 0.) / nodes
# max_pr = max(g, key = lambda n : n.rank).rank
max_pr = find_rank(g)
percentiles = [10, 20, 30, 40, 50, 60, 70, 80, 90, 95, 99]
perc = np.percentile(np.array(sorted(map(lambda n: n.rank, g))), percentiles)
pr_data.append(f'PR: InitialRank: {initial_rank}, Iteration: {iteration}, Percentiles {percentiles}: {perc}')
# print(f'InitialRank: {initial_rank}, Iteration: {iteration}, Avg. PR: {avg_pr}')
# print(f'InitialRank: {initial_rank}, Iteration: {iteration}, max PR: {max_pr}')
sr = sensor_rank(g)
avg_sr = reduce(lambda acc, n: acc + n.rank, sr, 0.) / nodes
# max_sr = max(sr, key = lambda n : n.rank).rank
max_sr = find_rank(sr)
perc = np.percentile(np.array(sorted(map(lambda n: n.rank, sr))), percentiles)
sr_data.append(f'SR: InitialRank: {initial_rank}, Iteration: {iteration}, Percentiles {percentiles}: {perc}')
# print(f'InitialRank: {initial_rank}, Iteration: {iteration}, Avg. SR: {avg_sr}')
# print(f'InitialRank: {initial_rank}, Iteration: {iteration}, max SR: {max_sr}')
table.append(f'{iteration+1} & {avg_pr:.8f} & {max_pr:.8f} & {avg_sr:.8f} & {max_sr:.8f} \\\\')
# out.write(f'{iteration+1} & {avg_pr:.8f} & {max_pr:.8f} & {avg_sr:.8f} & {max_sr:.8f} \\\\\n')
# with open(f'./{initial_rank}_{iteration+1}_pr.txt', 'w') as f:
# for n in g:
# f.write(f'{n.rank}' + '\n')
# with open(f'./{initial_rank}_{iteration+1}_sr.txt', 'w') as f:
# for n in sr:
# f.write(f'{n.rank}' + '\n')
avg_out = float(sum_out) / len(count_map.keys())
known_out = count_map[known_r]
print(f'\tavg_out: {avg_out}')
print(f'\tmin_out: {min_out}')
print(f'\tmax_out: {max_out}')
print(f'\tknown_out: {known} {known_out}')
out.write(f'bucket: {bucket}\n')
out.write(f'\tavg_out: {avg_out}\n')
out.write(f'\tmin_out: {min_out}\n')
out.write(f'\tmax_out: {max_out}\n')
out.write(f'\tknown_out: {known} {known_out}\n')
print()
print()
out.write('\n')
out.write('\n')
for row in table:
print(row)
out.write(f'{row}\n')
print()
print()
out.write('\n')
out.write('\n')
for row in sr_data:
print(row)
out.write(f'{row}\n')
print()
print()
out.write('\n')
out.write('\n')
for row in pr_data:
print(row)
out.write(f'{row}\n')
print()
print()
print()
print()
if __name__ == "__main__":
main()