Study on body fat density prediction

International Journal of Data Mining & Knowledge Management Process (IJDKP) Vol.5, No.3, May 2015
DOI : 10.5121/ijdkp.2015.5301 1
STUDY ON BODY FAT DENSITY PREDICTION
BASED ON ANTHROPOMETRIC VARIABLES
Shiva Shanth Reddy Ainala, Nawaf Aljohani, Kaushik Roy, Xiaohong Yuan, and
Huiming A. Yu
Department of Computer Science,
North Carolina State A&T University, Greensboro, USA
ABSTRACT
For a human body to function properly it is essential to have a certain amount of body fat. Fat serves to
manage body temperature, pads and protects the organs. Fat is the fundamental type of the body's vitality
stockpiling. It is important to have a healthy amount of body fat. Overabundance of fat quotient can build
danger of genuine wellbeing issues. Anthropometry is a broadly accessible and basic strategy for the
appraisal of body composition. Anthropometry measures are weight, height, Body Mass Index (BMI),
waist, boundary, biceps, skinfold etc. The human fat percentage is figured by taking anthropometric
variables. We proposed a methodology to determine the body fat percentage using R programming and
regression formula. We analyzed 10 anthropometric variables and 3 demographic variables. Our study
shows that the impact of certain variables has an edge over other in predicting body fat percentage.
KEYWORDS
Anthropometric variables, Body fat percentage, Fat prediction, Outliners using R programming.
1. INTRODUCTION
Fat is a manifestation of body tissue made out of cells which principally store lipids (unsaturated
fats and related mixes). These cells are embedded in a framework of connective tissue [1]. There
are two totally distinctive sorts of tissues: Brown Adipose Tissue (BAT) and White Adipose
Tissue (WAT) [1]. Fat is an important component for body, which supplies energy making
nutrients to function properly. Fat is source of essential fatty acids which are key dietary
requirements and provide energy [2]. Fat protects body organs from stun, keeps up temperature
of body and elevates solid cells to capacity [3]. Fat is imperative for generation of hormones in
our body [4]. Fat is additionally structure as defensive pads for indispensable organs, for
example, kidneys, heart and intestines [5]. Generally, if we look closely at the human body, there
are certain number of fat cells which remain more or less constant throughout our adult lives.
Therefore, we maintain an average ideal weight for the specific duration [1],[6].
Fat plays a vital role in our body mechanism. Specific measurements like anthropometric
measurements have been used for assessment of body composition [7]. Anthropometric variable
measurements such as Fat, Age, Height, Weight, Neck, Chest, Abdomen, Hip, Thigh, Knee,

2
Ankle, Extended biceps, Forearm and Wrist are taken as an important qualities to analyze the
body fat [8].
Anthropometry is a widely available and simple technique for the assessment of body
composition which does not need specialized laboratory or expensive equipment [5].
Anthropometry variables produce precise appraisal of fat percentage [6]. Anthropometric
measures are assessed for finding out the changes in individuals by valid methods [6], [9].
Various equations are generated to predict body fat percentage through anthropometric variables
in [8], [3], [10]. Diverse set of comparisons uses people weight, stature and sex, which include
smallest number of fat percentage predictions [11]. In fact, most of the previous methods measure
body density which is then used to estimate body fat percentage.
Therefore, previous methods do not measure body fat directly. As a result of that, the body fat
percentage depends on body’s weight. However, each tissue in individual’s boy has a different
density. Addressing the above issue, in this research effort, regression models are designed and
developed in an effort to predict the body fat percentage using 10 anthropometric variables which
have precise values [8]. The proposed models suggest that inexpensive and simple anthropometry
variables produces accurate estimate of percentage body fat [8], [3], [10]. This research study will
be helpful to determine if body fat could be predicted from anthropometric and demographic
variables. This study will also help identify how these variables are impacting on human body fat
percentage.
2. PROPOSED METHOD
Body fat can be divided into two categories essential fat which is necessary for normal healthy
functioning and storage fat which is the fat we accumulate beneath our skin, in certain specific
areas inside our body and in muscles. Based on an extensive literature review, the research
questions are developed. The experiments were conducted on a dataset that contains 4550
patient’s data in an attempt to distinguish the effective anthropometric and demographic variables
for assessing body fat based on the random sampling techniques [1]. The total sampling size is
4429. Data has been collected from “ROBA research Ltd” in India [8]. Different data analysis
techniques were used including data cleaning, data validation, correlation, regression, stepwise
regression, regression diagnostics. We adopted the best fit models for regression analysis.
The dependability and precision of estimation is compressed by connections in addition to t-tests
and regression. Regression equations were cross validated using simple demographic and
anthropometric measures. The equations were also utilized to estimate the fat rate. The basic
expansion in the equations would be the age and gender. In view of the experimental goals, we
proposed a general model with 10 anthropometric variables. Moreover, that model used
demographic data to estimate body fat percentage by using two models. Model 1 with 3 variables
and model 2 with all the variables. The body fat percentage measured through the proposed
models will help us monitor our body fat. This research effort also finds the variables which are
most effective in calculating body fat percentage.

3
Fig.1 Block diagram of Research Methodology
2.1 Data Analysis
R programming and R Rattle were used to analyze the data. The experiments were conducted
using Box plot and Histograms (for Normal distribution). Normality tests (Shapiro-wilk test)
were carried out for data reduction. Correlation and Stepwise Regression analysis were deployed
for estimating the ‘Cause-and-Effecting’ best fit model.
2.2 Data Access & Cleaning
All the Lab report data used for our analysis is available in the electronic format. By using
different techniques this data was converted into csv files. Total 4550 samples were converted
into data table format. Data cleaning and validation process was performed using R programming
Language.
2.3. Data Validation
We followed some inferential techniques for data normalization. The Box plot demonstrates the
outliers and histogram data shows the normal distribution. Shapiro-Wilk test was performed for
normality test.
The data used for analysis had more variation and deviations, and then data was normalized by
performing various data cleaning and data validation techniques. From Fig.2, it is found that dark
spots indicate a strong positive correlation between 0 to 1 while the lighter spots reveal weak
positive correlation between 0 to -1.

4
Fig.2 Data Correlation by Pearson Method
Fig. 3 Outliers identified using Boxplot

5
Fig.4 Normal distribution using Histograms
3. RESULTS
3.1 Test of Normality
From Figs. 3 and 4, we identified outliers for different anthropometry measures. These are Body
Fat (), Age (), Height (), Weight (), Neck (), Chest (), Abdomen (), Hip (), Thigh (), Knee (),
Ankle (), Extended biceps (), Forearm () and Wrist (). After removing outlier summarization, we
present the mean values and standard deviations (stds) in Table 1 for different anthropometry
measures.
3.2. Regression Model-I Analysis
From Table 2 we find that regression exploratory variables are 99% statistically significant.
Summary of regression correlation with R= 0.726, R Square = 0.527 and adjusted R Square =
0.526 shows a moderate modelling for predicting body fat. Age and Weight variables positively
predicted body fat. Height represents a negative predictor. Negative and positive exploration

6
variables are 99% statistically significant. Equation (1) regression Diagnosis. The values with this
equation closely predicts the fat percentage.
Body fat = 26.412+0.188*Age+0.543* Weight- 0.327*Height (1)
Table 1: SHAPIRO-WILK TEST FOR VARIABLES
3.3. Regression Model-II Analysis
From Table 3, we find that regression exploratory variables are 99% statistically significant.
Summary of regression correlation with R=0.843, R Square = 0.71 and adjusted R Square =
0.708 represents a very strong modeling for predicting body fac. Neck circ, Chest circ and Hip
circ variables negatively predict the body fat. Abdomen circumference, thigh circumference,
forearm circumference and wrist circumference variables have positiove impacts on predicting
the body fat. The other variables, including Knee circumference (cm), Ankle circumference (cm)
and extended biceps circumference (cm) are not statistically significant. Negative and positive
exploration variables are 95 % statistically significant in finding the body fat. Equation (2) is also
proposed for regression Diagnosis. This model has more precise values as compared to equation
(1).
Body FAT = 20.211-0.602*Neck circumference - 0.215* Chest circumference + 1.017*
Abdomen circumference (cm) - -0.417* Hip circumference + 0.203* Thigh circumference
+0.274* Forearm circumference +0.568*Wrist circumference (2)
variables Mean Std
Shapiro-Wilk
Statistic df Sig.
abdomen_circ 94.07 9.721 .994 1429 .000
Age 46.98 12.796 .988 1429 .000
ankle_circ 25.20 3.214 .992 1429 .000
body_fat 21.02 8.346 .997 1429 .003
chest_circ 102.54 7.924 .994 1429 .000
extended_biceps 34.38 4.041 .992 1429 .000
forearm_circ 30.85 3.492 .992 1429 .000
Height 180.81 7.076 .995 1429 .000
hip_circ 101.63 6.462 .997 1429 .005
knee_circ 40.73 3.617 .991 1429 .000
neck_circ 40.10 3.640 .980 1429 .000
thin_circ 61.39 5.412 .996 1429 .002
Weight 82.69 11.678 .992 1429 .000
wrist_circ 20.43 3.090 .970 1429 .000

7
Table 2: SHAPIRO-WILK TEST FOR REGRESSION MODEL 1
Table 3: SHAPIRO-WILK TEST FOR REGRESSION MODEL 2
4. CONCLUSION
This research provides an insight of how body fat can increase the risk of having serious health
issues such as high blood pressure, high cholesterol content, diabetes, cancer etc. Our results
conclude that the body fat percentage depends on different factors, including age, weight,
gender. All the anthropometric variables have their own impact while abdomen_circ, thigh_circ,
forearm_circ and wrist_circ have signification impacts on predicting body fat percentage. In
addition, this research also proposed regression models that can help predict the body fat
percentage. Therefore, based on anthropometric variables, one can determine the ideal value of
body fat percentage. The results we obtained provide a novel way of predicting body fat
percentage using specific anthropometric values.
REFERENCES
[1] P. L. ouchard C, Leblanc C, Tremblay A, Thériault G, "Inheritance of the amount and distribution of
human body fat," Int J Obes. 1988, vol. 12(3), pp. 205-215, 1988.
[2] Dollman, J., et al. (0000). "Dietary intake, physical activity and TV viewing as mediators of the
association of socioeconomic status with body composition: a cross-sectional analysis of Australian
youth." Int J Obes 31(1): 45-52.

8
[3] Jackson, A. S. and M. L. Pollock (1978). "Generalized equations for predicting body density of men."
British journal of nutrition 40(03): 497-504.
[4] Folsom, A. R., et al. (1990). "Incidence of hypertension and stroke in relation to body fat distribution
and other risk factors in older women." Stroke 21(5): 701-706.
[5] G. E. van der Ploeg, S. M. Gunn, R. T. Withers, and A. C. Modra, "Use of anthropometric variables
to predict relative body fat determined by a four-compartment body composition model," Eur J Clin
Nutr, vol. 57, pp. 1009-1016, //print 0000.
[6] C. Pongchaiyakul, Kosulwat, V., Rojroongwasinkul, N., Charoenkiatkul, S., Thepsuthammarat, K.,
Laopaiboon, M., Nguyen, T. V. and Rajatanavin, R, "Prediction of Percentage Body Fat in Rural Thai
Population Using Simple Anthropometric Measurements," Obesity Research, vol. 13, pp.729–738,
2005.
[7] Deurenberg, P., et al. (1991). "Body mass index as a measure of body fatness: age-and sex-specific
prediction formulas." British journal of nutrition 65(02): 105-114.
[8] Snehalatha, C., et al. (2003). "Cutoff values for normal anthropometric variables in Asian Indian
(Folsom, Prineas et al. 1990)adults." Diabetes care 26(5): 1380-1384.
[9] M. Ramirez-Zea, B. Torun, R. Martorell, and A. D. Stein, "Anthropometric predictors of body fat as
measured by hydrostatic weighing in Guatemalan adults," The American Journal of Clinical
Nutrition, vol. 83, pp. 795-802, April 1, 2006 2006.
[10] Jackson, A. S., et al. (1979). "Generalized equations for predicting body density of women."
Medicine and science in sports and exercise 12(3): 175-181.
[11] R. T. Withers, N. P. Craig, P. C. Bourdon, and K. I. Norton, "Relative body fat and anthropometric
prediction of body density of male athletes," European Journal of Applied Physiology and
Occupational Physiology, vol. 56, pp. 191-200, 1987/03/01 1987.

Study on body fat density prediction

More Related Content

What's hot (19)

Viewers also liked (20)

Similar to Study on body fat density prediction (20)

Recently uploaded (20)

Study on body fat density prediction