Back

Questionnaire based Prediction of Hypertension using Machine Learning

Chaturvedi, A.; Srivastava, S.; Rai, A.; Cheema, A. S.; Chelimela, D.; Aravindakshan, R.

2020-06-20 health informatics
10.1101/2020.06.18.20133397 medRxiv
Show abstract

Machine Learning has proven its ability in healthcare as an assisting technology for health care providers either by saving precious time or timely alerts or vitals monitoring. However, their application in real world is limited by availability of data. In this paper, we show that simple machine learning algorithms especially neural networks, if designed carefully, are extremely effective even with limited amount of data. Specifically with exhaustive experiments on standard Modified National Institute of Standards and Technology database (MNIST) dataset we analyse the impact of various parameters for effective performance. Further, on a custom dataset collected at a tertiary care hospital for hypertension analysis, we apply these design considerations to achieve better performance as compared to competitive baselines. On a real world dataset of only a few hundred patients, we show the effectiveness of these design choices and report an accuracy of 75% in determining whether a patient suffers from hypertension.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.