PREDICTING HOSPITAL STAY LENGTH USING KNN REGRESSOR OPTIMIZED WITH GRID SEARCH CV: AN EXPLAINABLE MACHINE LEARNING APPROACH

PREDICTING HOSPITAL STAY LENGTH USING KNN REGRESSOR OPTIMIZED WITH GRID SEARCH CV: AN EXPLAINABLE MACHINE LEARNING APPROACH

1 Navadeep Aluvoju, 2 Shaik Zaheer, 3 Keerthana Uppu, 4M V Anajana Devi

1,2,3UG Scholars,4 Associate Professor

1,2,3,4,Department of Computer Science and Engineering(Artificial Intelligence and Machine Learning)

1,2,3,4,Guru Nanak Institutions Technical Campus, Hyderabad, Telangana, India
--------------------------------------------------------***----------------------------------------------------------

Abstract
Predicting how long patients will stay in the ICU is truly critical to hospitals from a budgeting and cutting costs perspective. We considered applying a straightforward machine learning technique named K-Nearest Neighbors (KNN) in an effort to forecast ICU stays. The good thing about KNN is that physicians actually get to see how it is forecasted, as opposed to other sophisticated techniques which are a black box .

Our method operates by discovering similar patients for the current patient and predicting based on the duration they were hospitalized. We used an algorithm called GridSearchCV to discover the best parameters for our model. Through trials and combinations, we discovered that our method performs nearly as well as advanced algorithms but with a massive benefit - doctors can simply know why a prediction was made.

The findings indicate that straightforward, easy-to-communicate approaches are just as effective as advanced approaches in medicine. This is relevant because physicians must be confident and familiar with computer predictions before applying them for patient benefits. Our research indicates you do not necessarily require advanced algorithms to generate precise results.

Keywords: ICU length of stay, K-Nearest Neighbors, GridSearchCV, explainable AI, healthcare analytics, predictive modeling