Traditional mixed linear modelling versus modern machine learning to estimate cow individual feed intake
Three modelling approaches were used to estimate cow individual feed intake(FI) using feeding trial data from a research farm, including weekly recordingsof milk production and composition, live-weight, parity, and total FI.Additionally, weather data (temperature, humidity) were retrieved from theDutch National Weather Service (KNMI). The 2014 data (245 cows; 277parities) were used for model development. The first model (M1) applied anexisting formula to estimate energy requirement using parity, fat and proteincorrected milk, and live-weight, and assumed this requirement to be equal toenergy intake and thus FI. The second model used ‘traditional’ Mixed LinearRegression, first using the same variables as in M1 as fixed effects (MLR1), andthen by adding weather data (MLR2). The third model applied BoostedRegression Tree, a ‘modern’ machine learning technique, again once with thesame variables as M1 (BRT1), and once with weather information added(BRT2). All models were validated on 2015 data (155 cows; 165 parities) usingcorrelation between estimated and actual FI to evaluate performance. BothMLRs had very high correlations (0.91) between actual and estimated FI on 2014data, much higher than 0.46 for M1, and 0.73 for both BRTs. When validated on2015 data, correlations dropped to 0.71 for MLR1 and 0.72 for MLR2, andincreased to 0.71 for M1 and 0.76 for both BRTs. FI estimated by BRT1 was, onaverage, 0.35kg less (range: -7.61 – 13.32kg) than actual FI compared to 0.52kgless (range: -11.67 – 19.87kg) for M1. Adding weather data did not improve FIestimations.
| Main Authors: | , , , |
|---|---|
| Format: | Article in monograph or in proceedings biblioteca |
| Language: | English |
| Subjects: | Big Data, dairy cows, machine learning, precision feeding, prediction, |
| Online Access: | https://research.wur.nl/en/publications/traditional-mixed-linear-modelling-versus-modern-machine-learning |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Three modelling approaches were used to estimate cow individual feed intake(FI) using feeding trial data from a research farm, including weekly recordingsof milk production and composition, live-weight, parity, and total FI.Additionally, weather data (temperature, humidity) were retrieved from theDutch National Weather Service (KNMI). The 2014 data (245 cows; 277parities) were used for model development. The first model (M1) applied anexisting formula to estimate energy requirement using parity, fat and proteincorrected milk, and live-weight, and assumed this requirement to be equal toenergy intake and thus FI. The second model used ‘traditional’ Mixed LinearRegression, first using the same variables as in M1 as fixed effects (MLR1), andthen by adding weather data (MLR2). The third model applied BoostedRegression Tree, a ‘modern’ machine learning technique, again once with thesame variables as M1 (BRT1), and once with weather information added(BRT2). All models were validated on 2015 data (155 cows; 165 parities) usingcorrelation between estimated and actual FI to evaluate performance. BothMLRs had very high correlations (0.91) between actual and estimated FI on 2014data, much higher than 0.46 for M1, and 0.73 for both BRTs. When validated on2015 data, correlations dropped to 0.71 for MLR1 and 0.72 for MLR2, andincreased to 0.71 for M1 and 0.76 for both BRTs. FI estimated by BRT1 was, onaverage, 0.35kg less (range: -7.61 – 13.32kg) than actual FI compared to 0.52kgless (range: -11.67 – 19.87kg) for M1. Adding weather data did not improve FIestimations. |
|---|