Author(s): Bhavana Kamarthapu, Ajay Babu Kakani, Sri Krishna Kireeti Nandiraju, Sandeep Kumar Chundru, Srikanth Reddy Vangala and Ram Mohan Polam

Abstract: Diagnosis prediction involves using past patient visit information to Figure out what their future health might look like which is tough but key in healthcare informatics. This work sets up a machine learning (ML) framework that can be used for clinical prognosis by learning from a database of over 40,000 anonymized MIMIC-III ICU patients. The approach requires thorough handling of the data such as missing data treatment, excluding outliers, turning categories into numbers and standardizing all the features by Z-score normalization. Highlighted features are included and the data is separated into learning and testing groups. For classification problems involving huge data sets, KNN excels which is why it is used the model reaches an accuracy of 90%, precision and recall of 92%, and an F1-score of 92%. Comparative analysis with Restreet and ULM Fit models highlights the superior performance of the proposed approach, making it a reliable and efficient tool for clinical diagnosis and prognosis in intensive care settings.

DOI: 10.33545/27076571.2023.v4.i1a.170

Pages: 80-88 | Views: 500 | Downloads: 226

Download Full Article: Click Here