Paper Details

The clinical risk assessment processes involved in providing the highest level of preventive care for the patient have been established as a key/profound feature of modern healthcare systems, and with the continuing increase in availability of patient health data, the use of machine learning to predict clinical risk for patients has gained significant popularity among the medical community. This study provides a data-driven exploratory analysis of the use of machine-learning (ML) algorithms in order to develop a comparative evaluation of various algorithms for clinical risk assessment through exploring the data in EDA (Exploratory Data Analysis) prior to ML modeling via using a dataset of 2200 records created in CSVs (Comma Separated Values). EDA is performed on the dataset to investigate the variability of the values within the dataset, the characteristics of the dataset and the frequency distributions of the target classes prior to the modeling of the data. Multiple versions of ML algorithms were applied to the same 2200 record dataset and the evaluation was based on standard evaluation metrics. The results of this study indicate that by conducting an EDA on the dataset, the prediction accuracy and interpretability of the resulting ML models for clinical risk assessment were improved significantly.

Exploratory Data Analysis, Machine Learning, Clinical Risk Stratification, Classification, Predictive Analytics, Healthcare Data