diff options
| -rw-r--r-- | hawkes_experiments/cause.py | 57 |
1 files changed, 36 insertions, 21 deletions
diff --git a/hawkes_experiments/cause.py b/hawkes_experiments/cause.py index 17ec884..800f699 100644 --- a/hawkes_experiments/cause.py +++ b/hawkes_experiments/cause.py @@ -2,10 +2,9 @@ from cPickle import load from math import exp, sin from collections import Counter from csv import reader, writer -from data2 import parse +from data import parse import sys import networkx as nx -import matplotlib.pyplot as plt import numpy as np @@ -31,42 +30,58 @@ def cause(lamb, alpha, mu): fatal_droots += 1 continue background_rate = lamb * (1 + 0.43 * sin(0.0172 * t1 + 4.36)) - neighbors = sorted([(n2, t2, alpha / d * mu * exp(-mu * (t1 - t2))) - for (n2, t2, d) in s], reverse=True) - neighbor_rate = sum(e[2] for e in neighbors) + parents = sorted([(n2, t2, alpha / d * mu * exp(-mu * (t1 - t2)), d) + for (n2, t2, d) in s], reverse=True, + key=lambda x: x[2]) + parent_rate = sum(e[2] for e in parents) # if sum(e[2] for e in prl[:1]) > br: # G.add_edge((n1, t1), tuple(prl[0][:2])) - if background_rate > neighbor_rate: + if background_rate > parent_rate: roots += 1 if (n1, t1) in fatals: fatal_roots += 1 else: - G.add_edge((n1, t1), tuple(neighbors[0][:2])) + G.add_edge(tuple(parents[0][:2]), (n1, t1), + weight=parents[0][3]) # l.append(prl[0][2] / br) infections += 1 if (n1, t1) in fatals: fatal_infections += 1 - # l.sort(reverse=True) - # plt.plot(l) - # plt.show() return (droots, roots, infections, fatal_droots, fatal_roots, fatal_infections, G) def analyze_graph(G): - counts = Counter(len(c) for c in nx.weakly_connected_components(G)) - w = writer(open("components_dist.csv", "w")) - w.writerows(counts.most_common()) - edges = ((n1, t1, n2, t2) for ((n1, t1), (n2, t2)) in G.edges_iter()) - e = writer(open("edges.csv", "w")) - e.writerows(edges) + csizes = [len(c) for c in nx.weakly_connected_components(G)] + print "cascades: {0}, min: {1}, max: {2}, mean: {3}, median: {4}".format( + len(csizes), np.min(csizes), np.max(csizes), np.mean(csizes), + np.median(csizes)) + # counts = Counter(l) + # w = writer(open("components_dist.csv", "w")) + # w.writerows(counts.most_common()) + edges = list(G.edges_iter(data=True)) + print "edges: {0}".format(len(edges)) + times = [e[1][1] - e[0][1] for e in edges] + distances = [e[2]["weight"] for e in edges] + print "times, min: {0}, max: {1}, mean: {2}, median: {3}".format( + np.min(times), np.max(times), np.mean(times), np.median(times)) + print "distances, min: {0}, max: {1}, mean: {2}, median: {3}".format( + np.min(distances), np.max(distances), np.mean(distances), + np.median(distances)) + # e = writer(open("edges.csv", "w")) + # e.writerows(edges) if __name__ == "__main__": - nodes, edges, events, event_edges = load(open("data2.pickle", "rb")) - lamb, alpha, mu = 1.1847510744e-05, 0.00316718040144, 0.00393069204339 - # print len(event_edges), sum(len(e) for e in events.itervalues()) - # print len(fatal()) - (doors, roots, infections, fatal_droots, + nodes, edges, events, event_edges = load(open("data-dist1.pickle", "rb")) + lamb, alpha, mu = 1.86602117779e-05, 0.0433473674726, 0.00109325510695 + # lamb, alpha, mu = 1.87717287808e-05, 5.12006113875e+14, 4.20918377797e-20 + (droots, roots, infections, fatal_droots, fatal_roots, fatal_infections, G) = cause(lamb, alpha, mu) + r = "events: {0}, droots: {1}, roots: {2}, infections: {3}, "\ + "fatal_droots: {4}, fatal_roots: {5}, "\ + "fatal_infections: {6}".format(len(event_edges), droots, roots, + infections, fatal_droots, fatal_roots, + fatal_infections) + print r analyze_graph(G) |