2021/11/04 by Mina Samizadeh, Jessica C. Jones‐Smith, Samizadeh, Mina +5
Computer Science · #Artificial Intelligence (cs.AI) #Data Mining Algorithms and Applications #FOS: Computer and information sciences #Fuzzy Logic and Control Systems #Machine Learning (cs.LG) #Machine Learning and Data Classification
paper · pdf · doi:10.48550/arxiv.2111.04475
openalex publication_date 2021/11/04 · openalex created_date 2022/05/05 · openalex updated_date 2026/07/28
Overweight and obesity remain a major global public health concern and\nidentifying the individualized patterns that increase the risk of future weight\ngains has a crucial role in preventing obesity and numerous sub-sequent\ndiseases associated with obesity. In this work, we use a rule discovery method\nto study this problem, by presenting an approach that offers genuine\ninterpretability and concurrently optimizes the accuracy(being correct often)\nand support (applying to many samples) of the identified patterns.\nSpecifically, we extend an established subgroup-discovery method to generate\nthe desired rules of type X -> Y and show how top features can be extracted\nfrom the X side, functioning as the best predictors of Y. In our obesity\nproblem, X refers to the extracted features from very large and multi-site EHR\ndata, and Y indicates significant weight gains. Using our method, we also\nextensively compare the differences and inequities in patterns across 22 strata\ndetermined by the individual's gender, age, race, insurance type, neighborhood\ntype, and income level. Through extensive series of experiments, we show new\nand complementary findings regarding the predictors of future dangerous weight\ngains.\n