Visit our website: http://bit.ly/2GtXaiw
Data Balancing – A Problem Driven Case Study
Xiaonan Zhang, Senior Data Scientist, Johns Hopkins University Healthcare
Abstract:
Medical/Healthcare phenomena (diseases, death rate or other high utilizer prediction) always contribute to a small part of the population. Therefore, for any test of any models, there are trade-off between precision and recall, which is highly depended on how vital the phenomena are and how many resources are in the prevention healthcare. Here, we would use 30 days re-admission prediction as a problem-driven case study to discuss the mechanisms and some solutions behind this trade-off.
Add comment