Pemodelan Kenyamanan Sapi Perah Dara di Dataran Rendah Tropis Menggunakan Artificial Neural Networks Berdasarkan Laju Respirasi

Abstract

Lingkungan dataran rendah tropis umumnya memiliki suhu lingkungan/ambient temperature (AT) dan kelembapan relatif/relative humidity (RH) yang tinggi sehingga berpotensi menyebabkan heat stress pada sapi perah. Salah satu respons fisiologis yang sensitif terhadap perubahan kondisi lingkungan adalah laju respirasi/respiratory rate (RR), yang dapat digunakan sebagai indikator kenyamanan ternak. Penelitian ini bertujuan mengembangkan model artificial neural networks (ANN) untuk memprediksi RR sapi perah dara Friesian Holstein (FH) menggunakan data mikroklimat yang diperoleh melalui sistem internet of things (IoT) D-Ruminansia, serta mengklasifikasikan tingkat kenyamanan ternak berdasarkan nilai RR hasil prediksi. Penelitian dilaksanakan di Cibugary Dairy Farm, Jakarta Timur, menggunakan tiga ekor sapi perah dara FH berumur 12–15 bulan. Data AT dan RH dikumpulkan secara otomatis menggunakan perangkat D-Ruminansia, sedangkan RR diukur melalui pengamatan langsung. Analisis dilakukan melalui regresi linear sederhana untuk mengetahui hubungan antarvariabel, kemudian dilanjutkan dengan pemodelan ANN berarsitektur 2-4-1 dengan AT dan RH sebagai variabel input serta RR sebagai variabel output. Hasil penelitian menunjukkan bahwa kondisi mikroklimat kandang berada pada kisaran AT 24,5–37,2 °C dan RH 45,3–92,9%, sedangkan RR berada pada kisaran 28–72 rpm. Analisis regresi menunjukkan hubungan negatif yang sangat kuat antara AT dan RH (R2 = 0,98), hubungan positif antara AT dan RR (R2 = 0,98), serta hubungan negatif antara RH dan RR (R2 = 0,96). Model ANN menghasilkan nilai error sebesar 0,00046 dan menunjukkan hubungan yang sangat kuat antara RR aktual dan RR prediksi (R2 = 0,98), yang menunjukkan kesesuaian prediksi yang tinggi pada data penelitian. Simulasi model pada rentang AT 24–37 °C dan RH 45–93% menghasilkan prediksi RR sebesar 29–70 rpm. Berdasarkan kriteria klasifikasi yang digunakan, RR pada kisaran 26–50 rpm dikategorikan sebagai kondisi nyaman, sedangkan RR >50 rpm dikategorikan sebagai kondisi stres. Kondisi nyaman umumnya terjadi pada AT 24–27 °C, sedangkan transisi menuju kondisi stres mulai muncul pada kisaran AT 28–31 °C tergantung kombinasi RH. Pada AT 32–37 °C, sebagian besar kondisi lingkungan diklasifikasikan sebagai stres. Penelitian ini menunjukkan bahwa integrasi data mikroklimat D-Ruminansia dengan ANN dapat digunakan untuk memprediksi RR dan mengklasifikasikan tingkat kenyamanan sapi perah dara FH pada lingkungan dataran rendah tropis. Pendekatan ini berpotensi mendukung pengembangan sistem pemantauan kenyamanan ternak secara real-time berbasis IoT dan kecerdasan buatan.

Tropical lowland environments are generally characterized by high ambient temperature (AT) and relative humidity (RH), which may increase the risk of heat stress in dairy cattle. One of the physiological responses that is highly sensitive to environmental changes is respiratory rate (RR), which can be used as an indicator of animal comfort. This study aimed to develop an artificial neural networks (ANN) model to predict the RR of Friesian Holstein (FH) dairy heifers using microclimate data collected through the D-Ruminansia internet of things (IoT) system and to classify animal comfort levels based on the predicted RR values. The study was conducted at Cibugary Dairy Farm, East Jakarta, using three FH dairy heifers aged 12–15 months. Ambient temperature and relative humidity data were automatically collected using the D-Ruminansia device, while RR was measured through direct observation. Data analysis was carried out using simple linear regression to determine the relationships among variables, followed by ANN modeling with a 2-4-1 architecture, using AT and RH as input variables and RR as the output variable. The results showed that the barn microclimate conditions ranged from 24.5 to 37.2 °C for AT and from 45.3 to 92.9% for RH, while RR ranged from 28 to 72 respirations per minute. Regression analysis revealed a very strong negative relationship between AT and RH (R2 = 0.98), a strong positive relationship between AT and RR (R2 = 0.98), and a strong negative relationship between RH and RR (R2 = 0.96). The ANN model produced an error value of 0.00046 and demonstrated a very strong relationship between observed and predicted RR values (R2 = 0.98), indicating a high level of prediction agreement in the study data. Model simulations across AT and RH ranges of 24–37 °C and 45–93%, respectively, generated predicted RR values ranging from 29 to 70 respirations per minute. Based on the classification criteria used in this study, RR values ranging from 26 to 50 respirations per minute were classified as comfortable conditions, whereas RR values above 50 respirations per minute were classified as stressful conditions. Comfortable conditions generally occurred at AT ranging from 24 to 27 °C, while the transition toward stressful conditions began at AT ranging from 28 to 31 °C depending on RH combinations. At AT ranging from 32 to 37 °C, most environmental conditions were classified as stressful. This study demonstrates that the integration of D-Ruminansia microclimate data with ANN can be used to predict RR and classify the comfort levels of FH dairy heifers under tropical lowland conditions. This approach has the potential to support the development of real-time animal comfort monitoring systems based on IoT and artificial intelligence technologies.