about - adidotcom/Heavywaters-Project GitHub Wiki
Welcome to the Heavywaters-Project wiki! The idea of the project is to be able to successfully use machine learning cocepts and build a document classification model. A document labeled dataset is provided, that has 62204 entries of documents. Each entry has a document label and an OCR output of the document. All the words of the document was obscured using fixed length hashing.
The dataset looks something like this:
CANCELLATION NOTICE, 586242498a88 21e314d3afcc 818a7ff3bf29 4e43b72d46c0 578830762b27........ 43565b1afa44 5f6653c869fc
Step 1: Understanding the data The very first approach was to look at the data and understand what they mean at a very high level. The document can be labelled into 14 different categories and the following number of records:
BILL 18968
POLICY CHANGE 10627 CANCELLATION NOTICE 9731 BINDER 8973 DELETION OF INTEREST 4826 REINSTATEMENT NOTICE 4368 DECLARATION 968 CHANGE ENDORSEMENT 889 RETURNED CHECK 749 EXPIRATION NOTICE 734 NON-RENEWAL NOTICE 624 BILL BINDER 289 INTENT TO CANCEL NOTICE 229 APPLICATION 229