Sepsis Detection Using Extreme Gradient Boost (XGB): A Supervised Learning Approach
Asad Ullah, Huma Qayyum, Muhammad Khateeb Khan, Fawad Ahmad · 2021
Sepsis is one of the major trending topics in the field of Bio-Medical Sciences, Sepsis is a major disease, causing a lot of causalities, and a large amount of money is consumed for the diagnosis and treatment of sepsis found in patients. Early sepsis detection can decrease patients’ death rates and give a big economic relief to patients’ families. Many techniques have been used to detect sepsis earlier than clinical results are known, but machine learning approaches are leading one among all other techniques and tools. Many datasets are available which are used for more precise results, and some of the researchers use their own datasets, which are not easily available publicly. We used two datasets, Set A and Set B, for training and testing of our model, and these are publicly available at Physionet.org. We used Set A as a whole and a big proportion of Set B for the training of our algorithm and the remaining proportion of Set B for testing purpose [1]. We applied forward filling and backward filling on the dataset to fill the missing valuesand achieve the required training sets, then the extreme gradient boost (XGB) classifier is applied, which gives the result that sepsisis found. The sepsis is detection shows 92% accuracy. This algorithm would be used to increase patients’ chances of survival and would save the amount of money that is being used for diagnosis and treatment of sepsis annually.