# sorting the two vector by pid, but it does NOT make many sense...
# pid_vector, object_vector = zip(*sorted(zip(pids, system_objects))) # unpacking arguments by using `*`
for record in raw_dataset:
vector = list()
pid = record["pid"]
date = record["date"]
obj = record["object"]
vector.append(objects.index(obj)) # object index
vector.append(pids.index(pid)) # event pid index
vector.append(dates.index(date)) # date index
vector.append(system_objects[obj]) # number of system obj occurrence
vector.append(process_ids[pid]) # number of pid occurrence
vectorized_dataset.append(tuple(vector))
ksom = SOM(10, 10, vectorized_dataset, 100)
for iv in random.sample(vectorized_dataset, 100):
# print iv
# print objects[iv[0]], pids[iv[1]], dates[iv[2]]
ksom.epoch(iv)
pass
line_width = 10
for index, node in enumerate(ksom.nodes):
print ("(%d,%d):%d " % (node.x, node.y, node.bmu_count)),
if (index + 1) % line_width is 0:
print ""
pass
vector_dict['3'], # token index of activity
vector_dict['4'], # number of database operations
vector_dict['5'], # number of content provider queries
vector_dict['6'], # number of network operations
]'''
dataset.append(vector)
# dimension of the map (from config file)
config_file = open('./som_config.js')
configs = dict(map(lambda x:x[0:-1].split('='), config_file.read().strip().split('\n')))
config_values = configs.values()
width = int(config_values[0])
height = int(config_values[1])
# init the self-organising map
ksom = SOM(width, height, vector_dataset)
# start training
for activity_vector in random.sample(activity_dataset, len(activity_dataset)):
ksom.epoch(activity_vector['vector'], activity_vector['name'], activity_vector['start_date'])
pass
# get SOM nodes collection in database
collection = db.som_nodes
# store the map to database
for node in ksom.nodes:
map_node = dict()
map_node['x'] = node.x
map_node['y'] = node.y
map_node['features'] = node.weights_vector
map_node['extra_data'] = node.extra_data
