International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169 Volume: 6 Issue: 2 117 - 125 ______________________________________________________________________________________ 117 IJRITCC | February 2018, Available @ http://www.ijritcc.org _______________________________________________________________________________________ Artificial Neural Network: A New Approach for Prediction of Body Fat Percentage Using Anthropometry Data in Adult Females Sugandha Mehandiratta 1 Priyanka Singhal 2 A.K. Shukla 3 and Rita Singh Raghuvanshi 4 1 Ex-PG Student, Department of Mathematics, Statistics and Computer Science, College of Basic Sciences and Humanities 2 Ex-PG Student, Department of Foods & Nutrition, College of Home Science, 3 Dean, College of College of Basic Sciences and Humanities 4 Dean, College of College of Home Science (G. B. Pant University of Agriculture & Technology, Pantnagar, Uttarakhand, 263145, India.) Abstract: Assessing body fat using anthropometric data would be useful in predicting chronic diseases. Accurate use of proper statistical models in analysing body composition data is of prime importance. This study was undertaken to assess body composition of diseased and non-diseased women using body composition analyser thereafter using data for development of statistical model. The objective was to find relationship of various anthropometric parameters with Percent Body Fat (BF%) and to develop various prediction models for estimating BF on the basis of anthropometric data. BF% was predicted using Linear Regression (LR), Multiple Linear Regression (MLR), Non-Linear Regression (NLR) and Artificial Neural Network (ANN) models. The predictors used in the study were age (yrs.), height (cm), weight (kg), Body Mass Index (BMI) (Kg/m 2 ) and Waist Circumference (WC) (cm). Data utilized for the study was related to 860 adult females aged 18-60 years out of which 700 were non-diseased and 160 were diseased (diabetic and hypertensive). Out of various models developed using LR, MLR, NLR for Non-Diseased group, three predictors viz. age, BMI and WC were found to be appropriate for estimating BF%. However, the best prediction of BF% was achieved using ANN model taking age, height, weight and WC as predictors (R 2 = 0.787). ANN technique was found as the most suitable technique for developing prediction models for estimation of BF% in non-diseased group. However, in diseased group ANN model could not predict BF% more precisely, may be due to some other factors affecting the body composition of females of diseased group. Key Words: Body fat: Anthropometry: Artificial Neural Network: Women. __________________________________________________*****_________________________________________________ I. Introduction: Anthropometry is an important aspect used for understanding physical variation and assessing body composition. Despite the modern techniques, anthropometric measurements are important to study the genetic structure and prediction of risk factors of many complex diseases in human health. Anthropometric parameters such as height, weight, body mass index, and waist circumference evaluate underweight and obesity conditions. Anthropometric measurements are easy to conduct in field studies and have better acceptance in the communities. Overall obesity and visceral obesity in which a high proportion of body fat is deposited on the trunk and in the abdomen are associated with deleterious health outcome. The prevalence of obesity is increasing rapidly worldwide. The causes of obesity are not completely understood but according to heredity and decreased physical activity are its major causes (1) . It is observed that reduction in lean body mass and relative increase of percentage body fat mass is due to the age related change in chemical composition of an individual (2) . Age is associated with regional fat distribution causes negative health consequences. Body fat gets redistributed with age. Visceral obesity is predicted outcome of increased waist circumference. Waist circumference rises with advancing age. It has been reported that among older group larger waist circumference than that of younger one (3) . An increased risk of developing cardiovascular disease, chronic disorders and disabilities was determined in the case of females with a WC over 80 cm, and a strongly increased risk in the case of females with a WC of over 88 cm was observed (4) . Mean value for fat (%) was observed as 19.0±4.4 and 29.2±5.6 cm at WC cut off points viz., <80 cm and ≥80 cm respectively. It shows presence of higher body fat in abdominally obese female adults (5) . The use of statistical models in medical diagnoses and biomedical research may affect individual’s life by predicting whether their health is protected or is in danger. Therefore, careful and accurate use of proper statistical models in analysing health related data is of prime importance. BF% is the percent of total body weight of a person. BF % can be predicted from BMI (6) . The ideal weight and BF% varies for males and females of different age groups. Excess body fat contributes to an array of medical diseases and can greatly increase the risk of contracting conditions such as coronary heart disease, diabetes, cancers, gallstones etc. Deurenberg et al developed prediction equations based on anthropometric parameters for both males and females using LR technique. It was also