High Blood Pressure Classification Using Meta-heuristic Based Data-Centric Hybrid Machine Learning Model

Abstract

Blood Pressure (BP) is created when the heart pumps the blood into blood vessels. If this pressure is more than 140/90 mmHg, it is diagnosed as high blood pressure (HBP). If HBP is not noticed and treated at an early level, it may lead to life-threatening issues. The design and development of machine learning models (MLM), to predict HBP in advance based on bio-psychological factors is gaining the attention of people. MLM is assisting medical doctors in diagnosing diseases more accurately, though MLM is exceptionally doing great in this domain, they are data-dependent. Conventional MLM is evaluated for the considered dataset. The major pitfall of such a model is a dependency on the dataset. If the same model is exposed to different datasets of the same type, the performance of the model may not be consistent. This paper proposed a heuristic-based dynamic data-drive, Age, Anger level (AA)-anxiety level, cholesterol level, obesity level (ACO) based hybrid MLM to predict HBP. The proposed model initially calculates the degree of dependency in terms of Pearson correlation between the attributes and class label attributes. The model is said to be hybrid as it uses the correlation-driven apriori-based fuzzy association rule miner to predict HBP. The proposed approach is data-centric and dynamic, as it calculates the Pearson correlation value for the given dataset at runtime and also assigns the priority value to the attribute at run time. The experimental setup is done on 1100 data records; the proposed model has got 91.168% accuracy, precision of 0.946, and recall of 0.933. The output of the model is a fuzzy inference engine consisting of the top 10 meta-heuristic-based fuzzy association rules, these rules can be used by a person as a knowledge base to manage and treat HBP.