benchmark result

add_bulk_bench.py

 -*- coding:utf-8 -*-                                                                                                                                           

import jubatus
from jubatus.common import Datum

import random
import time

cl = jubatus.Anomaly("localhost", 9199, "test", 0)
cl.clear()
random.seed(1)
datum_list = []
datum_length = 100
datum_num = 3000
for i in range(datum_num):
    d = Datum()
    for x in range(datum_length):
        d.add_number("{}".format(x), random.random())
    datum_list.append(d)

start = time.time()
cl.add_bulk(datum_list)
end = time.time()
print("add_bulk time:", end - start)
print(cl.calc_score(datum_list[-1]))

cl.clear()
start = time.time()
for d in datum_list:
    cl.add(d)
end = time.time()
print("add time:", end - start)
print(cl.calc_score(datum_list[-1]))

light_lof.json

{
  "converter" : {
    "string_filter_types" : {},
    "string_filter_rules" : [],
    "num_filter_types" : {},
    "num_filter_rules" : [],
    "string_types" : {},
    "string_rules" : [
      { "key" : "*", "type" : "str", "sample_weight" : "bin", "global_weight" : "bin" }
    ],
    "num_types" : {},
    "num_rules" : [
      { "key" : "*", "type" : "num" }
    ]
  },
  "parameter" : {
    "nearest_neighbor_num" : 10,
    "reverse_nearest_neighbor_num" : 30,
    "method" : "lsh",
    "parameter" : {
      "hash_num" : 64
    }
  },
  "method" : "light_lof"
}

lof.json

{
  "converter" : {
    "string_filter_types" : {},
    "string_filter_rules" : [],
    "num_filter_types" : {},
    "num_filter_rules" : [],
    "string_types" : {},
    "string_rules" : [
      { "key" : "*", "type" : "str", "sample_weight" : "bin", "global_weight" : "bin" }
    ],
    "num_types" : {},
    "num_rules" : [
      { "key" : "*", "type" : "num" }
    ]
  },
  "parameter" : {
    "nearest_neighbor_num" : 10,
    "reverse_nearest_neighbor_num" : 30,
    "method" : "lsh",
    "parameter" : {
      "hash_num" : 64
    }
  },
  "method" : "lof"
}

result.txt

$ jubaanomaly -f lof.json &
$ python add_bulk_bench.py
2017-12-11 16:46:35,288 27231 INFO  [anomaly_serv.cpp:246] model cleared: 
add_bulk time: 3.411644458770752
1.0249063968658447
2017-12-11 16:46:39,061 27231 INFO  [anomaly_serv.cpp:246] model cleared: 
add time: 18.150259017944336
1.0471123456954956

$ jubaanomaly -f light_lof.json &
$ python add_bulk_bench.py
2017-12-11 16:47:54,393 27290 INFO  [anomaly_serv.cpp:246] model cleared: 
add_bulk time: 0.7454001903533936
1.033468246459961
2017-12-11 16:47:55,480 27290 INFO  [anomaly_serv.cpp:246] model cleared: 
add time: 2.91017746925354
1.033468246459961

I have updated the benchmark result after applying jubatus/jubatus_core@a3784d4