def display_vector_index_details(vector_index, vectors, vector_names, vectorizer):
# borrow the feature -> string method from get_common_features_from_cluster
output = ""
upf, unf, shared_features = get_common_features_from_cluster(vectors[vector_index, :].reshape(1, -1), np.array([0]), vectorizer)
ip = vector_names[vector_index]
features = set()
for feature in reduce_shared_features(shared_features[0]['positive']):
features.add(" - ".join([i.encode("unicode-escape") for i in feature]))
features = sorted(list(features))
output += "IP: {0}\n".format(ip)
for f in features:
output += f + "\n"
return output + "\n"
def display_shared_vector_indeces_details(vector_indeces, vectors, vector_names, vectorizer):
# borrow the feature -> string method from get_common_features_from_cluster
output = ""
upf, unf, shared_features = get_common_features_from_cluster(vectors[vector_indeces, :], np.array([0] * len(vector_indeces)), vectorizer)
features = set()
for feature in reduce_shared_features(shared_features[0]['positive']):
features.add(" - ".join([i.encode("unicode-escape") for i in feature]))
features = sorted(list(features))
if len(features) == 0:
return "No shared features\n"
for f in features:
output += f + "\n"
return output + "\n"
def display_vector_index_details(vector_index, vectors, vector_names,
vectorizer):
# borrow the feature -> string method from get_common_features_from_cluster
output = ""
upf, unf, shared_features = get_common_features_from_cluster(
vectors[vector_index, :].reshape(1, -1), np.array([0]), vectorizer)
ip = vector_names[vector_index]
features = set()
for feature in reduce_shared_features(shared_features[0]['positive']):
features.add(" - ".join([i.encode("unicode-escape") for i in feature]))
features = sorted(list(features))
output += "IP: {0}\n".format(ip)
for f in features:
output += f + "\n"
return output + "\n"
def display_shared_vector_indeces_details(vector_indeces, vectors,
vector_names, vectorizer):
# borrow the feature -> string method from get_common_features_from_cluster
output = ""
upf, unf, shared_features = get_common_features_from_cluster(
vectors[vector_indeces, :], np.array([0] * len(vector_indeces)),
vectorizer)
features = set()
for feature in reduce_shared_features(shared_features[0]['positive']):
features.add(" - ".join([i.encode("unicode-escape") for i in feature]))
features = sorted(list(features))
if len(features) == 0:
return "No shared features\n"
for f in features:
output += f + "\n"
return output + "\n"
normalized_vectors = normalize(vectors)
reduced_vectors = pca(vectors)
# Cluster the vectors
logging.info("Clustering")
labels = cluster(vector_names, vectors, reduced_vectors,
normalized_vectors, vectorizer, args.strategy,
args.method, args.n_clusters, args.epsilon,
args.min_samples, args.metric)
logging.info("Clustering Complete")
# Test cluster validity
overall_score, per_cluster_score = validate_clusters(vectors, labels)
# Analysis relevant to the person reading results
universal_positive_features, universal_negative_features, shared_features = get_common_features_from_cluster(
vectors, labels, vectorizer)
# Reduce results and relevant information to per cluster data
cluster_details = {}
for cluster_id in per_cluster_score.keys():
cluster_details[cluster_id] = {
"silhouette":
per_cluster_score[cluster_id],
"shared_positive_features":
shared_features[cluster_id]['positive'],
#"shared_negative_features": shared_features[cluster_id]['negative'],
"ips": [
vector_names[x] for x in xrange(len(vector_names))
if labels[x] == cluster_id
]
}
logging.info("Vectorizing complete")
logging.info("Reducing vector dimensions with PCA")
normalized_vectors = normalize(vectors)
reduced_vectors = pca(vectors)
# Cluster the vectors
logging.info("Clustering")
labels = cluster(vector_names, vectors, reduced_vectors, normalized_vectors, vectorizer, args.strategy, args.method, args.n_clusters, args.epsilon, args.min_samples, args.metric)
logging.info("Clustering Complete")
# Test cluster validity
overall_score, per_cluster_score = validate_clusters(vectors, labels)
# Analysis relevant to the person reading results
universal_positive_features, universal_negative_features, shared_features = get_common_features_from_cluster(vectors, labels, vectorizer)
# Reduce results and relevant information to per cluster data
cluster_details = {}
for cluster_id in per_cluster_score.keys():
cluster_details[cluster_id] = {
"silhouette": per_cluster_score[cluster_id],
"shared_positive_features": shared_features[cluster_id]['positive'],
#"shared_negative_features": shared_features[cluster_id]['negative'],
"ips": [vector_names[x] for x in xrange(len(vector_names)) if labels[x] == cluster_id]
}
print_cluster_details(cluster_details, shared_features)
if args.plot:
create_plot(reduced_vectors, labels, vector_names)
