Generate insights from diabetes dataset using deviation-based approach

Dataset: https://archive.ics.uci.edu/ml/datasets/Diabetes+130-US+hospitals+for+years+1999-2008

This result using diabetes clean version dataset: see https://github.com/rischanlab/Cleaning_diabetes_130_US_hospital_dataset

Example target and reference query that we used, 4 aggregate functions are used (avg, max, sum, count):

Target query: select A, M(F) from diabetes where readmitted = 'NO' group by A (subset readmitted = NO)

Reference query: select A, M(F) from diabetes group by A (whole dataset)

Top-10 insights as shown in Figure below

Example plot target and reference views from top-k

num_emergency means emergency visits in the year before the hospitalization

Target view I

Reference view I

Target view 2

Reference view 2

Target view 3

Reference view 3