h_anomaly

A small example package Anomaly detection using hierarchical clustering, anomaly detector, classifiers and fast model rebuilding

Required Packages to run -

1) pandas
2) numpy
3) pickle
4) scipy
5) freediscovery
6) sklearn
7) matplotlib

Install the Package

pip install -U h_anomaly

To Use the Package:

Import h_anomaly

import h_anomaly - 'from h_anomaly import driver'

Building Tree for the 1st time

cluster,cluster_tree,max_depth = driver.cluster_driver(file_path,target_class,default_class)

Loading New data for testing

df,train_X,train_y = driver.get_data(file_path,target_class,default_class)

Storing Test Data for future uses

test_df,test_X,test_y = driver.get_data(file_path,target_class,default_class)
cluster.set_test(test_X,test_y)

Certify Model for performance monitoring:

cluster.certify_model(cluster_tree,test_y)

Check Cluster model for retraining

cluster.check_model(cluster_tree,threshold)

Available Functions:

1) fit - Fit the Data into the Birch algorithm to create the clusters
  def fit(self,data,y)
2) set_test - Store the test data for future uses
  set_test(self,data,y)
3) get_cluster_tree - For each cluster at every level creates the bcluster objects
  get_cluster_tree(self)
4) model_adder - Classification model added to each cluster by this function (Change this function to add different model)
  def model_adder(self,cluster_tree)
5) update_model - Classification model is updated with new data
   update_model(self,cluster_tree,cluster_id)
6) outlier_model_adder - Outlier detection model is added to each cluster (Change this function to add different model)
  outlier_model_adder(self,cluster_tree)
7) certify_model - Scores are calculated in this function
  certify_model(self,cluster_tree,test_y)