Artificial Intelligence for Debt Collection
Sirit is an engine for assisting customers with multichannel digital interaction. Our automated processes can segment, apply strategies and learn from past experiences to improve future outcomes.
Sirit hires us to develop a Machine Learning model to predict the best strategy to collect debt from a customer.
Debt collection agencies have to deal with large portfolios of debts from various creditors. Managing and prioritizing these portfolios efficiently can be a complex task, affecting the overall effectiveness of collections.
We used several strategies to find useful information from the data:
- Data profiling
- Clustering (we’re in love with UMAP)
- Classification
We also prepared a detailed report with the results and the steps to follow to improve the model. Our idea is to use AI to improve the collection process and to help the agents to be more efficient in these main areas:
- AI can help debt collection agencies in optimizing their portfolios by providing insights into the most effective strategies for different types of debtors.
- AI can detect early signs of potential payment issues and flag accounts that are likely to become delinquent. This allows collectors to intervene proactively and prevent further escalation.
- AI algorithms can learn from each interaction and improve over time. This continuous learning enables collectors to refine their strategies and enhance their effectiveness.
- AI can assist in identifying fraudulent activities and distinguishing between genuine debtors and those attempting to evade payments.
Technologies in this project:
- Python, pandas, scikit-learn, seaborn, matplotlib, jupyter notebooks.
- Mito and pandas profiler.
- UMAP for clustering.