Improving Farm Management Information Systems with Data Mining

Abstract

Over the past several years, farm enterprises have grown in size substantially while their number has steadily declined. As the size of their farms grow more and more farmers are deploying information systems, commonly called as Farm Management Information Systems (FMIS), to manage the day to day activities of their farms. The deployment of FMIS enable farmers to capture detailed data that can potentially be analysed by data mining tools to provide valuable information for optimizing the farm enterprises. However, data mining is generally not a common feature of many FMIS. In order to evaluate the suitability of data mining for use in FMIS, two case studies were performed using data captured in FMIS and applying various data mining algorithms. Microsoft Azure Machine Learning Studio is chosen because it provides a simple drag-and-drop visual interface that can be used by farm domain experts. In this study, two common problems were addressed in dairy farming: calving prediction of dairy cows and prediction of lactation value of milking cows. In both cases data mining models were built and experiments were run and results in both cases indicate that the required data is available from FMIS and data mining techniques provides acceptable performance. It was also shown that farm domain experts can easily use a user-friendly and drag-and-drop data mining tools with minimal initial training. Based on the insight from the two case studies and literature study, several decision problems that can be addressed with data mining such as heat prediction and lameness prediction were identified.