Optimizing Ongole Grade Cattle weight prediction through Principal Component Analysis

Abstract

Abstract Ongole Grade (OG) cattle are commonly raised by many farmers in Indonesia, particularly in rural and remote areas where weighing scales are not readily available. The weight of these OG cattle can be estimated by utilizing their body measurements through the application of Principal Component Analysis (PCA). This research aimed to predict the body weight (BW) of OG cattle, which are primarily kept by smallholder farmers, using PCA based on various body measurements such as body length (BL), chest girth (CG), shoulder height (SH), and chest width (CW). Additionally, the effectiveness of PCA-based predictions was compared with a multiple linear regression model. This study involved a total of 120 OG cattle, comprising 26 males and 94 females. The PCA of the body measurements, as well as the correlation and regression between these measurements (BL, CG, SH, and CW) and BW, were analyzed using the R programming language. The selection criteria for identifying the best-fit model for BW prediction were based on statistical indicators including coefficient of determination (R2), Adjusted R2, residual standard error (RSE), Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC). The findings of this investigation revealed that the primary factors representing body measurements were PC1 for both male (accounting for 89.21% variance) and female OG cattle (accounting for 85.71% variance). In comparison to the regression models, those generated from three PCs for males and two PCs for females demonstrated greater precision and simplicity in estimating the BW of OG cattle, without encountering multicollinearity issues. Consequently, the outcomes of this study have practical applications, serving as a means to predict the BW of OG cattle and contributing to selection programs within this cattle breed.