Obtaining accurate individual feed intake records is a key first step in achieving genetic progress toward a more efficient pig for nutrient utilization. Feed intake records collected by electronic feeding systems contain errors (extreme values or outliers), which are due to feeder malfunction or animal movements. In this study, we introduce a new feed intake data editing strategy to replace errors and missing observations occurring in feed intake data, based on multiple imputation methods. Compared with the well-established linear mixed model (LMM) approach, multiple imputation either by using conditional distribution (MI) or by chained equation (MICE) results in increased accuracy of data adjustment in simulated phenotypes with artificially introduced errors. Feeder visit records in the simulated data sets were sampled from a data set including individual pig feed intake visits collected by Smithfield Premium Genetics from year 2004 to 2013. Three scenarios were considered in the analysis with 5%, 10% and 20% error visits simulated. Each scenario was replicated 5 times. Accuracy of the error-adjustments was measured as correlation between the true error-free daily feed intake (DFI) or average daily feed intake (ADFI), and the adjusted ones. Multiple imputation methods outperformed the linear mixed model approach in all scenarios with average accuracies of 96.71%, 93.45% and 90.24% obtained with MI and 96.84%, 94.42% and 90.13% obtained with MICE, compared with 91.0%, 82.63% and 68.69% using LMM for DFI with simulated error rate 5%, 10% and 20%, respectively. Similar results were obtained for ADFI. In conclusion, multiple imputation was introduced in this study as a more accurate and flexible error-adjustment method for feed intake data collected by electronic feeders.