CASPR
1.0.0
CASPR is a transformer-based framework for deep learning from sequential data in tabular format, most common in business applications.
Tasks critical to enterprise profitability, such as customer churn prediction, fraudulent account detection or customer lifetime value estimation, are often tackled by models trained on features engineered from customer data in tabular format. Application-specific feature engineering however adds development, operationalization and maintenance costs over time. Recent advances in representation learning present an opportunity to simplify and generalize feature engineering across applications.
With CASPR we propose a novel approach to encode sequential data in tabular format (e.g., customer transactions, purchase history and other interactions) into a generic representation of a subject's (e.g., customer's) association with the business. We evaluate these embeddings as features to train multiple models spanning a variety of applications (see: paper). CASPR, Customer Activity Sequence-based Prediction and Representation, applies transformer architecture to encode activity sequences to improve model performance and avoid bespoke feature engineering across applications. Our experiments at scale validate CASPR for both small and large enterprise applications.
CASPR: Customer Activity Sequence-based Prediction and Representation (NeurIPS 2022, New Orleans: Tabular Representation Learning)
Build
python==3.9, setuptools
python setup.py build bdist_wheel
Installation
(now)
pip install .distAI.Models.CASPR-<ver>.whl[<optional-env-modifier>]
(future)
pip install AI.Models.CASPR[<optional-env-modifier>]
use any of below modifiers, to customize the installation for target system / usecase:
horovod - for distributed training and inference on Horovod
databricks - for distributed training and inference on Databricks
aml - for (distributed) training and inference on Azure ML
hdi - for execution on Azure HD Insights
xai - to enable explainability
test - for extended test execution
dev - for development purposes only
Examples
(TODO: can we point to a well commented one of our examples w/ or w/o data?)
We welcome contributions! Please see the contribution guidelines.
For feature requests or bug reports please file a GitHub Issue.
This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact [email protected] with any additional questions or comments.
This project is licensed under the MIT License.
