Table 1.
Case definition of different classification terminologies used.
Case definition of different classification of cases, CCT (Corrective Claw Trimming) was done to animals which had physiological hoof growth, TCT (Therapeutic Claw Trimming was done to animals suffering from hoof disease), SP (Study personnel) one of the study personnel(SP) Performed the locomotion scoring.
Fig 1.
Distribution of daily behaviors and rolling averages of 7-day slopes of these behaviors for cows subjected to TCT and healthy control.
Fig 2.
Distribution of daily behaviors and rolling averages of 7-day slopes of these behaviors for cows subjected to CCT and healthy control.
Fig 3.
Distribution of daily behaviors and rolling averages of 7-day slopes of these behaviors for cows suffering from infectious lameness and non-infectious lameness.
Fig 4.
Distribution of daily behaviors and rolling averages of 7-day slopes of these behaviors for cows affected with moderate lameness and severe lameness conditions.
Fig 5.
Representation of model performance scores of ML models used for classification of cows needing hoof trimming vs healthy.
Conventional feature model consisted of three variables i.e., daily lying time, daily steps and daily change; the Slope feature model included 7 day rolling deviation of each of these variables; All feature models included variables included in Conventional feature and Slope features.
Table 2.
Best performing model scores among the features tested (Conventional feature consisting of daily lying, daily steps and daily change recorded from sensor, slope feature consisting of a 7-day deviation each feature in Conventional feature, all features consisting of those in Conventional feature and slope feature) for each machine learning models used namely, ROCKET classifier, Random Forest, Naïve bayes, and Logistic Regression.
Fig 6.
The representation of model performance scores of ML models used for classification of cows needing therapeutic correction vs healthy.
Conventional feature model consisted of the three accelerometer derived variables i.e., daily lying, daily steps and daily change; Slope feature model included 7 day rolling deviation of each of these variables; All feature model included variables included in Conventional feature and Slope feature.
Fig 7.
The performance of different algorithms for classification of cows whether they have infectious lameness or non-infectious lameness condition among the diseased cows.
Conventional feature model consisted of three accelerometer derived variables i.e., daily lying, daily steps, and daily change; Slope feature model included 7 day rolling deviation of each of these variables; All feature model included variables included in Conventional feature and slope feature.
Fig 8.
The performance of different algorithms for classification of cows whether they have infectious lameness or non-infectious lameness condition among the diseased cows.
Conventional feature model consisted of the three accelerometer derived variables i.e., daily lying, daily steps and daily change; Slope feature model included 7 day rolling deviation of each of these variables; All feature model included variables included in Conventional feature and Slope feature.