Cleansing data from an electronic feeding station to improve estimation of feed efficiency

Abstract

Electronic feeding stations (EFS) are frequently used to automatically record feeding behaviour in pigs, but the resulting data often require cleansing. In this study, several filters were used to cleanse the data from an EFS on a commercial farm, spanning a period of 2147 days (2,709,600 records or visits, corresponding to 2748 pigs). The raw data of each animal was filtered based on three main parameters: fattening duration, feed ingested and live mass. Anomalies that affected EFS readings included aspects related to animal handling (e.g., changes of pen, deaths; affecting 26.1%), maintenance failures (e.g., obstructions, feed mass calibration; affecting 4.4%) and malfunctioning of the animal weighing scale (e.g., calibration, anomalies in the load cell; affecting 10.6%). A detailed analysis of the mass curves of each animal helped identify and categorise three typical errors: 1) large fluctuations of animal mass over a short period of time (affecting 7.3%); 2) poor calibration of the animal weighing scale (affecting 3.3%); and 3) absence of data from the weighing scale (affecting 1.9%). Once the raw data were filtered, feed efficiency was calculated, resulting in a mean value of 0.434, which is consistent with the literature. After filtering the first dataset, another dataset was used from the same farm to validate the results, with similar findings. Overall, the filtering methodology used helped to automate data analysis from EFS and improved the estimation of feeding efficiencies for individual animals. • EFS are frequently used to automatically record feed intake and mass of pigs. • An automated data cleansing process was developed to eliminate anomalous EFS data. • Anomalies were related to animal handling and weighing scale malfunctioning. • Data cleansing process produced reliable individual feed efficiency estimations.