US based collections company sought to improve the scheduling of the collections process to save costs and improve recovery.

Experience with recovery data and the associated problems leveraged our ability to work around the many data issues facing the client.

Built multiple Machine Learning models for various sub-processes as follows:

Conducting grid search and train algorithms with deep learning

Applied the trained model to produce prediction statistics on test data

Produce prioritised variable significance from learned model

Critically analysed the ML model accuracy through backtesting

Finally built an AI scheduling agent that assists the clients decision making

The client has a better way to assess the expected return and associated resourcing of their operation.

The AI algorithms are set up to continue to be learning in production, and will improve over time.