AmaleshV

use_tpu = True
if use_tpu:
    # Create distribution strategy (TPU initialization)
    tpu = tf.distribute.cluster_resolver.TPUClusterResolver()
    tf.config.experimental_connect_to_cluster(tpu)
    tf.tpu.experimental.initialize_tpu_system(tpu)
    strategy = tf.distribute.experimental.TPUStrategy(tpu)
    # Create model
    with strategy.scope():
        model = create_model()

AmaleshV

def create_model():
    # Name of the xlnet model to use
    model_name = 'xlnet-large-cased'
    # Max length of tokens
    max_length = 10
    # Load transformers config and set output_hidden_states to False
    config = XLNetConfig.from_pretrained(model_name)
    config.output_hidden_states = False
    # Load xlnet tokenizer
    tokenizer = XLNetTokenizerFast.from_pretrained(pretrained_model_name_or_path = model_name, config = config)

AmaleshV

def create_model():
    # Name of the xlnet model to use
    model_name = 'xlnet-large-cased'
    # Max length of tokens
    max_length = 10
    # Load transformers config and set output_hidden_states to False
    config = XLNetConfig.from_pretrained(model_name)
    config.output_hidden_states = False
    # Load xlnet tokenizer

AmaleshV

from googletrans import Translator
import time
import ast
from random_user_agent.params import SoftwareName, OperatingSystem
software_names = [SoftwareName.CHROME.value]
operating_systems = [OperatingSystem.WINDOWS.value, OperatingSystem.LINUX.value]   
user_agent_rotator = UserAgent(software_names=software_names, operating_systems=operating_systems, limit=200)
user_agent2 = user_agent_rotator.get_random_user_agent()
def split_list(alist, wanted_parts): #split the list to a wanted parts
    length = len(alist)

AmaleshV

from faker.config import AVAILABLE_LOCALES
#filtering AVAILABLE_LOCALES
country=['ar_EG','fakerar_PS','ar_SA','bg_BG','bs_BA','cs_CZ','de_DE','dk_DK','el_GR'
,'en_AU','en_CA','en_GB','en_NZ','en_US','es_ES','es_MX','et_EE','fa_IR','fi_FI'
,'fr_FR','hi_IN','hr_HR','hu_HU','hy_AM','it_IT','ja_JP','ka_GE','ko_KR','lt_LT',
'lv_LV','ne_NP','nl_NL','no_NO','pl_PL','pt_BR','pt_PT','ro_RO','ru_RU','sl_SI','sv_SE','tr_TR','uk_UA','zh_CN','zh_TW']
from tqdm import tqdm
from faker import Faker
import csv
def datagenerate(records, headers):

AmaleshV

https://colab.research.google.com/drive/1ln_OoicKGP23ysUmPiivJup2xWXIk5u9#scrollTo=Tf0ICF3O9vL5

AmaleshV

instance_to_explain=44
### Automative Narrative using SHAP for Transaction Monitoring###
final_ts['Alert_description_shap']=""
for instance_to_explain in range(0,50):
    
    column_names_explain=final_ts.columns #column names
    obs_instance_to_explain=final_ts.iloc[instance_to_explain,]
    obs_instance_to_explain_shapvalues =np.around(shap_values[1][instance_to_explain],3)
    prob_score_SAR_obs_instance_to_explain= ytest_rf[instance_to_explain,1]

AmaleshV

mcd = create_model('mcd',fraction=outlier_fraction)

AmaleshV

sod = create_model('sod',fraction=outlier_fraction)

AmaleshV

abod = create_model('abod',fraction=outlier_fraction,verbose=True)