Designing a decision-making tool for predicting hypertension with machine learning methods in Azar cohort population

Abstract

Abstract Background: This study was conducted in northwest Iran with the aim of predicting high blood pressure in adults aged 35 and older, and data mining algorithms were used to identify common patterns and risk factors related to blood pressure. Objective: The present study was conducted with the aim of designing a decision-making tool for predicting blood pressure with machine learning methods in the population of Azar cohort. Materials and methods: The present study is of an applied developmental type and this study was carried out in 1401-1403 with the aim of designing a decision-making tool for predicting blood pressure with machine learning methods based on the data set obtained from The Azar cohort, which included 14,984 records, was performed. This study was conducted in two stages of development and evaluation, using data mining technologies such as key influencer, what-if, clustering and decision tree. The effectiveness of the system was evaluated using established criteria, including accuracy and area under the ROC curve. Results: An operational prediction model was created to evaluate the possibility of hypertension in the population of AZAR group. The results showed that the important factors in increasing blood pressure are age, abdominal obesity, body mass index, physical activity and depression. The most common factors in the clustering of patients were gender, smoking and blood pressure status, based on which patients can be classified into different categories. A decision tree with 82 nodes was generated, the results showed that age is the main factor in predicting blood pressure through if-then statements, the accuracy was 0.822, and the area under the curve was 0.77. Conclusion: Conclusion: Developing a questionnaire to diagnose the disease and using a predictive model for blood pressure can be effective in identifying people at risk and applying preventive interventions.

Key words: Data Mining, Machine Learning, Hypertension, Prediction, Cohort