Qingyuan Yu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Qingyuan Yu has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
- Jiamei Song
College of Animal Science and Technology, Northeast Agricultural University, Harbin 150030, China
1 shared publication Liang Wang
Honghe University
1 shared publicationTingyi Hou
Northeast Agricultural University
1 shared publicationWeizheng Shen
Northeast Agricultural University
1 shared publicationXiaoli Wei
Northeast Agricultural University
1 shared publicationYonggen Zhang
Northeast Agricultural University
1 shared publication
Latest Publications
- Open Access
- 1 January 2022
- Published by Elsevier in SSRN Electronic Journal
Predicting Milk Yield Using Long Short-Term Memory Based on Wavelet Analysis of the Data
Milk is an important production indicator for dairy farms. Accurate prediction of milk yield can evaluate cow production performance and provide a reference for ranchers to make decisions. Here, we continuously monitored the milk output of a dairy farm for one year, exclude the cows with diseases and large parity, and screened out the milk yield data of 100 cows. It was found that the milk production was time-dependent, and the lactation curve was non-stationary. We designed a prediction model combining wavelet transform and Long Short-Term Memory Neural Network (WT-LSTM). The model first denoises the lactation curve, and then inputs it into the LSTM neural network for prediction, and further verifies the denoising effect through the prediction indicators. The results shown that in predicting the daily output, the performance of the sym7 mother wavelet denoising model is better than that of the control group without noise reduction, in which RMSE, MAPE, and MAE are reduced by 20%, 17%, and 22%, respectively, and ρ is increased by 5%. The wavelet transform can solve the problem of non-stationary milk quantity and improve the prediction performance of LSTM. In the actual production of dairy cows, adding wavelet transform to predict milk quantity can improve pasture management.
- 1 January 2022
- Published by Elsevier in SSRN Electronic Journal
Open AccessPredicting Milk Yield Using Long Short-Term Memory Based on Wavelet Analysis of the Data
Milk is an important production indicator for dairy farms. Accurate prediction of milk yield can evaluate cow production performance and provide a reference for ranchers to make decisions. Here, we continuously monitored the milk output of a dairy farm for one year, exclude the cows with diseases and large parity, and screened out the milk yield data of 100 cows. It was found that the milk production was time-dependent, and the lactation curve was non-stationary. We designed a prediction model combining wavelet transform and Long Short-Term Memory Neural Network (WT-LSTM). The model first denoises the lactation curve, and then inputs it into the LSTM neural network for prediction, and further verifies the denoising effect through the prediction indicators. The results shown that in predicting the daily output, the performance of the sym7 mother wavelet denoising model is better than that of the control group without noise reduction, in which RMSE, MAPE, and MAE are reduced by 20%, 17%, and 22%, respectively, and ρ is increased by 5%. The wavelet transform can solve the problem of non-stationary milk quantity and improve the prediction performance of LSTM. In the actual production of dairy cows, adding wavelet transform to predict milk quantity can improve pasture management.