Computerized-system and method for generating a reduced size superior labeled training dataset for a high-accuracy machine learning classification model for extreme class imbalance of instances

A computerized-system for generating a reduced size superior labeled training dataset for a high-accuracy machine learning classification model for extreme class imbalance, said computerized-system comprising:

• a database of majority labeled class instances;
• a database of minority labeled class instances;
• a superior labeled training database for a machine learning classification model;
• a memory to store the database of majority labeled class instances, the database of minority labeled class instances;
• and the superior labeled training database;
• and
• a processor, said processor is configured to:
• (a) retrieve instances from the database of minority labeled class instances and a preconfigured amount of instances from the database of majority labeled class instances to mark the retrieved instances as related to an initial dataset;
• (b) retrieve a pool-based sample of instances from the majority labeled class instances database;
• (c) select an instance to operate a clustering classification model on the selected instance and the instances that are marked as related to the initial dataset to yield clusters;
• (d) operate a learner model to:
• i. measure each instance in the yielded clusters according to a differentiability estimator and an indicativeness estimator;
• ii. mark measured instances as related to an intermediate training dataset according to a predefined measurement of the differentiability estimator and a predefined measurement of the indicativeness estimator;
• (e) repeat steps (c) through (d) until a preconfigured condition is met;
• (f) apply a variation estimator on all marked instances as related to an intermediate training dataset to select most distant instances;
• (g) mark the selected most distant instances as related to a superior training dataset;
• (h) check if the preconfigured condition is met, if not mark the superior training dataset as related to the initial dataset and repeat steps (b) through (h) until the preconfigured condition is met, if yes proceed to step (i);
• (i) store the instances marked as related to a superior training dataset in the superior labeled training database as part of a reduced size superior labeled training dataset.