Development of an expert system for the evaluation of reproductive performance and management of Virginia dairy farms
An expert system for dairy herd reproductive management for microcomputer was developed using an expert system shell and Turbo Pascal. A dairy extension reproductive specialist provided information for the system and empirical support was provided by research. The expert system initially examines days open, days to first insemination, percent of possible estruses observed, and number of breedings per conception to determine whether a problem exists. Interpretations ranging from “excellent” to “severe” were established for each parameter. “Excellent” and “adequate” interpretations correspond to a 12 to 13 mo calving interval. The system then selects for evaluation one of three areas that influences days open; days to first insemination, efficiency of detection of estrus, or conception percentage. Once an area has been selected for further evaluation, the expert system utilizes information from the user and from DHIA reproductive management reports developed by the Dairy Records Processing Center in Raleigh, NC. The reproductive reports are captured in a computer file and read by the expert system to identify problems of conception categorized by production, parity, service, days in milk, breed, and service sire. In addition, questions are asked by the expert system to isolate problems in data accuracy, semen handling, AI technique, detection of estrus, signs of estrus, and other management areas. Recommendations and suggestions are given. The expert system was designed to be used by extension personnel who may not have extensive knowledge of computers or reproductive management. The compiled program runs on an IBM compatible personal computer with 640K memory. Ten Virginia DHIA herds with conception problems were evaluated by the expert system and the extension specialist. Of 100 potential problem areas, the expert system and extension specialist identified 47, agreeing on 85% of them. Most discrepancies resulted from the expert applying a more restrictive standard when values were close to a preselected threshold.