Identification of Key Biometric Risk Dimensions Related to Obesity and Blood Pressure: A Pca-Based Study

Obesity is a global health concern which leads to a lot of non-communicable diseases such as hypertension, diabetes, dyslipidemia, sleep disorders, cancers, and cardiovascular diseases by impacting mortality. [4] Hypertension is persistent elevation of arterial pressure and a major risk factor for heart, brain, kidney and other diseases. [8] The 6 variables are selected related to obesity and blood pressure in this study. The obesity measurements are weight, height, waist circumference and hip circumference. The blood pressure measurements are mean systolic blood pressure and mean diastolic blood pressure readings. The secondary data of 501300 women in India was used in this study. The well-known dimension reduction technique called Principal Component Analysis (PCA) is utilized in this study to identify the two biometric risk indices from these variables. PCA is applied in several fields for identifying key dimensions of health data such as dietary patterns, cardiovascular risk calculation and biometric data. [2][3][6] As a pre-requisite to PCA, Bartlett’s test and Kaiser-Meyer-Olkin (KMO) test were performed to check the correlation criteria and sampling adequacy. [10] The variance of the data is maximized in PCA and 1st PC is worked on the entire variance of the data set. [9] The extractions of variance to the 2nd PC by variables weight, height, waist circumference, hip circumference, mean systolic blood pressure reading and mean diastolic blood pressure reading are 81.7%, 21.2%, 79.6%, 82.6%, 84.7% and 85.3% respectively. In this study. the six principal components are extracted 48.315%, 24.209%, 15.330%, 4.640%, 4.390% 3.115% of variance respectively. The graphical representation of eigen values versus component numbers is called as scree plot which commonly practiced to identify the number of extracted principal components. [7] The 70% of cumulative variance is exceeded by 2nd PC and the eigen value is greater than one only for first two principal components in this study. Therefore these 6 variables can be compressed by using 2 principal components. The interpretation of extracted principal components can be well explained by rotated component matrix. Moreover, the variables represented relatively high values of this rotated component matrix are key factors. In this study, PC 1 represents obesity data and PC 2 represents blood pressure measurements. It can be verified by the data point clusters of component plot in rotated space. Eventually, these two key biometric risk dimensions can be applied to model the health risk.