The use of integer programming in dairy sire selection.

Abstract

A computerized sire selection integer program, IPSIRE, was developed to test the optimality of solutions obtained by using MAXBULL, a popular computerized sire selection aid. The IPSIRE program uses an optimization software package and a compatible spreadsheet; MAXBULL uses linear programming techniques. The computer programs were compared using large and small data sets. Multiple combinations were run, varying the levels requested for PD fat percentage, PD type, and price paid per unit of semen. All other constraints were held constant. The objective function for all combinations maximized PD milk. The IPSIRE program exceeded MAXBULL by an average of $2.5 for PD dollars and 7.3 kg (16.1 lb) for PD milk using the small bull data base. Sire averages for traits PD type, PD fat percentage, and price paid were not different between the two programs. Similar trends were seen using the large bull data base. The IPSIRE program exceeded MAXBULL by an average of $1.2 for PD dollars and 4.1 kg (9.0 lb) for PD milk, whereas sire averages for PD fat percentage, PD type, and price paid were not different between programs. The goal for PD type had the least influence of the three goals on sire averages for PD milk. The goal for price paid had the greatest impact on realized PD milk; higher goals for price paid produced solutions with higher values for realized PD milk. The IPSIRE program required longer computational times for finding solutions that maximized PD milk than MAXBULL.