A Study on Prediction Model of the Hypertension Risk by Deep Learning


  • Sung-Jun Kim
  • Dea-Woo Park


Background/Objectives: Recently, as the research on the neural network deep learning based on machine learning technique is active, there are increasing attempts to apply this technology in various industries. In particular, studies on the prediction of hypertension risk in the healthcare sector had a limitation that the questionnaire and cohort DB were not used to accurately reflect the hypertension risk prediction.
Methods/Statistical analysis: In this study, we will use Python to collect public health data for 2018, and develop a model to predict the risk of hypertension and what are the main factors affecting the development of hypertension using deep learning analysis.
For the study, Four deep learning analyzes of decision tree, random forest, gradient boosting and logistic regression were used, and among these, logistic regression showed the best prediction rate. So, in this paper, we analyzed the risk factors for hypertension using logistic regression methodology.
Findings: As a result, age, women, BMI, height, weight, waistline, systolic blood pressure, diastolic blood pressure, urinary protein, serum creatinine, and gamma-tiffy variables were associated with high risk of hypertension. And the logistic regression showed the highest predicted value of 79.41%, and showed the concordance values for each disease variable affecting the prediction. The regression equation for the risk of hypertension is Log (p / (1-p)) =-54.9372 + 0.5228 (60s) -0.1752 (Famale) -0.1652 (BMI) -0.0549 (height) +0.0502 (weight) +0.0204 (waist) Circumference) +0.2601 (constrictor blood pressure) +0.2634 (diastolic blood pressure) +0.2272 (urea protein) +0.2271 (cholesterol) -0.0029 (gamma tipi).
Improvements/Applications: As a result, if the regression formula is used to produce a program for predicting hypertension in the future, it is possible to provide a service for self-diagnosis of hypertension risk.