Algorithms for identifying errors in individual feed intake data of growing pigs in group-housing

The present study describes algorithms for identifying errors in feed intake data of pigs, recorded with singlespace computerized feeding stations. Potential causes of errors are failed identification of pigs or an incorrect recording of feeder weight or time by the feeding station. Feed intake data of 250 pigs, divided into 30 groups, were analyzed. Data contained 385,329 records on visits of which 0.95% had no identification. Nine algorithms were developed to check data for errors caused by incorrect recordings. Algorithms focused on feed intake per visit, feeding rate per visit, or on the similarity of recorded feeder weights of subsequent visits. By using all nine algorithms, 6% of the visits were classified as being incorrect. The numbers of errors needs to be kept small, as it is impossible to adjust feed intake data without bias. Results indicated several instances where a feeding station functioned sub-optimally during a period of days or weeks. Frequent checking and correction of a feeding station function during recording would, therefore, reduce these errors. Expanding a feeding station’s software with the editing system described herein would allow a daily check of recorded data for errors. Furthermore, frequent maintenance of feeding stations will probably reduce the number of incorrect recordings.