The Health Resources & Services Administration’s (HRSA) grant management system activity page features a lengthy left-hand menu containing an extensive list of activities. Users often find themselves confused and frustrated as they scroll through this exhaustive list to locate their desired activity. Similarly, the EHB comprises over 300 reports, each with comprehensive metadata. New users typically need to examine the provided metadata to determine the purpose, search criteria, and result set of each report, making it challenging to find the specific report they require.
REI analyzed 18 months of system logs and transactional data to train machine learning algorithms, enabling the prediction of individual user-based activities and report preferences. The REI team employed user-based and rank-based collaborative filtering algorithms to identify the best-fit model. In the resulting user-activity matrix, each row represents a user, while each column signifies an activity. A dot product analysis was utilized to generate recommendations for the relative fit (likelihood of selection) for each activity and user combination. This Activity matrix could then be employed in real-time to suggest activities to display or highlight for each user, enhancing their experience.
The automation of activity and report selection led to an improved user experience for the HRSA grant management system. This enhancement not only boosted the productivity of staff members but also promoted the effective utilization of available reports, ultimately benefiting the overall system performance.
- Deep Learning
- Model Optimization
- Systems Automation
- Predictive Activity Modeling