Body Fat Percentage (BFP) is a measurement of total body fat that is used as an accurate measurement for the diagnosis of obesity. BFP measurement is sometimes difficult and inconvenient to perform, even though the picture of BFP’s value is very important for someone to find out the chances of being obese. To overcome this, data mining techniques can be used to measure the predictions of BFP values in a more practical way. This study implements data mining techniques, namely the CART and M5’ algorithm to predict a person's BFP value based on his/her body measurement. The CART algorithm uses the sample average values at leaf nodes to make numerical predictions, while the M5' algorithm builds a regression model for each leaf node with a hybrid approach. Regression trees provide a simple way of explaining the relationship between features and numerical results, but more complex model trees also provide more accurate results. In this study, the results show that the M5' algorithm is superior to the BFP dataset with a correlation value of 0.86 and an MAE value of 3.86.
Copyrights © 2023