Dairypert: An Expert Systems Approach for Improving Dairy Farm Management Practices and Assisting Technology Transfer
Kalter, Robert J.; Skidmore, Andrew L.
This manuscript reports on the development an expert system (DairyPert™) for application to production dairy operations in the United States. The approach combines rule and model based components in a system that permits diagnostic evaluation of a dairy's current operations. Neuron Data's NEXPERT shell is used for implementation. The shell is integrated with FoxPro database programs for data entry and delivery of system findings and results, and with an Excel spreadsheet containing the Cornell Net Carbohydrate and Protein nutrition model. Both IBM and Apple Macintosh personal computers are used as platforms for the effort. The knowledge base is structured modularly to determine constraints to greater profitability and/or production in eight key management areas. The eight key areas are nutrition (including feed ration evaluations), physical facilities, herd health, reproduction, replacement, milking practices, herd management ability, and economic constraints (including risk factors). Heuristics or rules of thumb used by experts in these fields are used in conjunction with the nutrition model for diagnostic purposes. For a number of the areas, input data, rules and system solution values are tracked separately for each of three facility types (freestall, corral or stanchion), for each physical enclosure or pen comprising the dairy operation, and for up to fourteen physiological animal classes within each enclosure. Thus, solution results are specified to a level of detail that permits the tendering of focused advice or management suggestions. Other key areas are concerned with problems of management and economics that relate to the overall herd situation. During system operation, results are stored at the level of detail discussed above. However, recognizing that merely listing the findings could lead to information overkill, a second set of rules are used by the system to prioritize and tailor the advice provided. The principle objective is to match the type of advice given to the management ability and economic capacity of the farm operator, as well as to the expected economic payoff and payback period. The resulting software package, using data for any dairy farm currently under an appropriate record keeping system, is capable of diagnosing dairy herd management problems and providing advice to the herd operator on how to correct such problems in a manner that will improve efficiency and production. The overall goal is to improve the short-and long-run management of and profitability from the dairy farm business and to prepare individual dairy operations for the adoption of new technology.
A.E. Res. 91-9
Charles H. Dyson School of Applied Economics and Management, Cornell University