Pemodelan Durasi Berbaring Sapi Perah Menggunakan Artificial Neural Network di Dataran Rendah Cibugary Farm

Abstract

Produktivitas sapi perah dipengaruhi oleh kondisi lingkungan kandang, terutama suhu dan kelembapan relatif yang tinggi pada wilayah tropis dataran rendah, yang dapat meningkatkan beban panas serta memengaruhi respons fisiologis dan perilaku ternak. Lama berbaring merupakan salah satu indikator perilaku untuk mengevaluasi respons sapi terhadap perubahan lingkungan, namun hubungan antara mikroklimat dan perilaku tersebut bersifat kompleks serta tidak selalu linier. Oleh karena itu, penelitian ini bertujuan menganalisis kondisi mikroklimat kandang, mengidentifikasi pola perilaku berbaring sapi perah, mengevaluasi kemampuan Artificial Neural Network (ANN) dalam memprediksi lama berbaring berdasarkan kondisi mikroklimat, serta mengkaji keterkaitan hasil prediksi dengan respons fisiologis dan manajemen pemeliharaan. Penelitian dilaksanakan di Cibugary Farm, Jakarta Timur, pada Desember 2025 hingga Januari 2026 menggunakan lima ekor sapi Friesian Holstein (FH) laktasi kedua. Suhu dan kelembapan relatif kandang dipantau selama 24 jam menggunakan sistem D-Ruminansia, sedangkan perilaku berbaring diamati selama tujuh hari menggunakan kamera closed-circuit television (CCTV) dan dianalisis menggunakan perangkat lunak CowLog 3.0.2. Pengukuran laju respirasi dilakukan sebagai indikator respons fisiologis terhadap perubahan lingkungan, sedangkan data suhu, kelembapan, dan perilaku berbaring digunakan untuk membangun model ANN berarsitektur 2-4-1. Hasil penelitian menunjukkan bahwa mikroklimat kandang berfluktuasi mengikuti pola harian, dengan suhu meningkat pada siang hari dan kelembapan relatif meningkat pada malam hingga dini hari. Durasi berbaring sapi lebih tinggi pada malam hingga dini hari dan menurun pada pagi hingga siang hari, yang dipengaruhi oleh perubahan mikroklimat maupun aktivitas rutin kandang seperti pemerahan, pemberian pakan, dan pembersihan kandang. Model ANN mampu memprediksi durasi perilaku berbaring berdasarkan kondisi mikroklimat dengan tingkat ketelitian yang baik. Hasil prediksi menunjukkan bahwa peningkatan suhu dan kelembapan relatif diikuti oleh peningkatan durasi berbaring, namun kondisi tersebut tidak selalu mencerminkan tingkat kenyamanan yang lebih baik. Pengukuran laju respirasi menunjukkan bahwa pada beberapa kondisi suhu tinggi sapi telah meningkatkan aktivitas termoregulasi, sehingga peningkatan lama berbaring lebih menggambarkan respons adaptif terhadap beban panas. Penelitian ini menunjukkan bahwa ANN mampu menggambarkan hubungan nonlinier antara suhu, kelembapan relatif, dan lama berbaring sapi perah. Integrasi data mikroklimat berbasis D-Ruminansia dengan pemodelan ANN berpotensi mendukung pengembangan sistem pemantauan perilaku sapi perah yang lebih objektif dan adaptif dalam pengelolaan lingkungan kandang serta evaluasi kesejahteraan ternak.

Dairy cow productivity is influenced by barn environmental conditions, particularly high ambient temperatures and relative humidity in tropical lowland areas, which increase heat load and affect the animals' physiological and behavioral responses. Lying duration is one of the behavioral indicators used to evaluate cows' responses to environmental changes. However, the relationship between barn microclimate and lying behavior is complex and does not always follow a linear pattern. Therefore, this study aimed to analyze barn microclimate conditions, identify daily lying behavior patterns in dairy cows, evaluate the performance of an Artificial Neural Network (ANN) in predicting lying duration based on microclimate conditions, and examine the relationship among predicted results, biological reactions, and herd management practices. The study was conducted at Cibugary Farm in East Jakarta from December 2025 to January 2026, using five second-lactation Friesian Holstein (FH) cows. Ambient temperature and relative humidity were continuously monitored for 24 hours using the D-Ruminansia system, while lying behavior was observed for seven consecutive days using closed-circuit television (CCTV) cameras and analyzed with CowLog 3.0.2 software. Respiration rate was measured as an indicator of biological reactions to environmental variations. Temperature, relative humidity, and lying behavior data were then used to develop an ANN model with a 2-4-1 architecture. The results showed that the barn microclimate exhibited a distinct daily pattern, with ambient temperature increasing during the daytime and relative humidity increasing from nighttime to early morning. Dairy cows spent more time lying during the night and early morning, while lying duration decreased from morning to midday. These behavioral patterns were influenced not only by changes in the barn microclimate but also by routine management practices, including milking, feeding, and barn cleaning. The ANN model reliably predicted lying duration based on barn microclimate conditions with good predictive accuracy. The prediction results indicated that increases in ambient temperature and relative humidity were associated with longer lying duration; however, this condition did not necessarily signify improved animal comfort. Respiration rate measurements proved that under several elevated-temperature conditions, the cows had already activated thermoregulatory responses, suggesting that the increased lying duration represented an adaptive response to heat load rather than improved welfare. This study shows that ANN can model the nonlinear relationships among ambient temperature, relative humidity, and dairy cow lying duration. The combination of D-Ruminansia-based microclimate data with ANN modeling has the potential to support the design of a more objective and adaptive behavioral monitoring system.