Investigating mineral affecting alkaline phosphatase activity and feed efficiency of broiler chickens using meta analytical and machine learning approaches

Articles in Press

Document Type : Research Paper

Authors

Department of Poultry Science, Faculty of Agriculture Science, Tarbiat Modares University, Tehran, Iran

Abstract

This study aims to develop two separate models to predict serum alkaline phosphatase (ALP) activity in broiler chickens based on dietary mineral intake and to forecast feed conversion ratio (FCR) based on their mineral intake and serum ALP activity. A meta-analysis method was employed to aggregate data from 29 articles spanning years 1998 to 2020. This resulted in a dataset of 185 rows containing variables such as serum ALP activity, calcium, phosphorus, zinc of the diet, and FCR in broiler chickens. Machine learning techniques, specifically artificial neural network models (ANN), were utilized for data analysis and predictive modeling. The ANN demonstrated high accuracy in predicting ALP activity and FCR, achieving R2 values of 97% and 95%, respectively. Sensitivity analysis revealed that serum ALP activity is more responsive to changes in calcium, whereas FCR is more sensitive to changes in zinc. Furthermore, through optimization of the ANN model, the minimum attainable FCR was found to be 1.41. This corresponded to ALP activity of 1190 U/L, and daily intake of zinc of 11.21 mg, phosphorus of 0.46 g, and calcium of 0.70 g. These findings provide insights into optimizing broiler chicken nutrition for improved performance. The developed model not only accurately predicts ALP activity and FCR in broiler chickens but also enhances broiler breeding by offering an easy-to-use tool for optimizing mineral intake and accurately predicting bird performance. To facilitate accessibility for readers and nutritionists, an Excel® calculator was created for predicting ALP activity and FCR in broilers using the developed ANN.

Keywords

Main Subjects

Iranian Journal of animal Science

Articles in Press, Accepted Manuscript
Available Online from 10 July 2024

Files

History

  • Receive Date: 30 November 2023
  • Revise Date: 27 June 2024
  • Accept Date: 29 June 2024

Share

How to cite

Statistics