agbocsardi.com / rtpm_assignment_2
[READ-ONLY] Mirror of https://github.com/agbocsardi/rtpm_assignment_2.
Star
Fork 0
Watch
Jupyter Notebook 100%
30 1 0
Part 2-3 & general train-test-split
Part 2-4
Preprocessing
.DS_Store
.gitignore
README.md
README.md

Categories #

declined = A_denined
cancelled = A_cancelled
approved = A_pending

RTPM cite this Professional 1 for the event classification !WINNER! Professional 4

RTPM cite this Academic 1 for a different event classification & Section 6 Analysis of Incomplete Cases

RTPM cite this !WINNER! Academic: Academic 2 for a different filtering cases not yet finished - section 2 last part - might prove above papers' event clf are the same

Plan #

  1. get all the Application_IDs out of each category above
  2. then build new, trace-based dataframe

Trace DF #

  1. We take every application process one by one
  2. From each we extract the full event-trace, and prefixes
    1. when it comes to prefixes, only stuff after A_accepted are interesting, everything before that is the same
  3. Figure out the features to use

Init Features - event logs #

case_attr = ['FirstWithdrawalAmount', 'NumberOfTerms', 'Accepted', 'MonthlyCost', 'Selected', 'CreditScore', 'OfferedAmount', 'OfferID', 'case:LoanGoal', 'case:ApplicationType', 'case:concept:name', 'case:RequestedAmount']

event_attr_cat = ['org:resource', 'concept:name', 'lifecycle:transition']
event_attr_num = ['time:timestamp']

TODO #

How to encode datetime in pandas df?