As we seen in medical environment that the patient is required overnight care (in-care) or not (out-care) after a medical procedure or surgery by analysing his diagnostic information. In this process, the medical staff assess the vitals of the patient (like age, gender, haemoglobins, erythrocyte etc.) and then decide whether the patient requires an overnight observation in the hospital or he is free to go home. Precision is seen to be a superior criterion to assess a model's effectiveness in data science applications in the medical field.It is so, because when it comes to human health, it is essential to have an algorithm with low False Negative Rate. By analysing Precision (instead of Accuracy) we ensure the same. Although the algorithms mentioned above perform fairly in terms of accuracy, they don’t compare precision and recall as the metric for performance. When contrasting the effectiveness of the classification models, we developed a unique step of a probability-based decision support system in this article.The new DSS provides better results because of the varying threshold on probability. This provides an extra edge to improve precision as well as recall so that overall performance of the model is improved by minimal loss to accuracy.

Machine learning, Classification, Confusion Matrix, Ensemble, Stacking, DSS, Neural Network