2026-10-15T11:11:06Z https://data.mendeley.com/oai

oai:data.mendeley.com/kxvsv72sbj.1 2026-09-01T13:11:40Z

Sang, Gaoli Electronic Ear-Tag Temperature and Estrus Event Records Mendeley Data This dataset contains longitudinal electronic ear-tag records and manually recorded estrus-event labels used in a retrospective sow estrus classification study. The data were obtained from routine production monitoring at one commercial swine farm in Zhejiang, China. Low-power electronic ear tags associated each observation with an individual sow and timestamp and transmitted the records through a wireless gateway to a local server. The retrospective analysis introduced no experimental intervention or change to normal animal management. The file sow_ear_tag_monitoring_records.xlsx contains 2,663,644 raw monitoring records from 409 sows collected between 8 November 2024 and 7 July 2025. Because of the Excel row limit, the records are divided across three worksheets with the same eight-column schema: sEarTagCode (electronic ear-tag identifier), sSowsNo (sow identifier/ear number), sBrand (farm-recorded breed or line), dBreedDate (mating date), dWeanDate (weaning date), tLastUploadTime (measurement upload timestamp), iTemperature (ear temperature, °C), and iStep (cumulative step-count reading). Sampling intervals are non-uniform. The file estrus_event_labels_and_sow_ids.pdf contains 521 positive sow-interval entries organized by date and twice-daily observation interval (08:00–09:00 and 15:00–16:00). Farm personnel performed manual estrus checks at approximately 08:30 and 15:30. The listed interval therefore represents the recorded detection time rather than the exact biological onset of estrus. Sow identifiers correspond to sSowsNo; numeric identifiers should be handled as strings. Not every labeled entry had sufficient matching temperature history to generate a retained analytical window. For the associated study, valid ear-temperature observations were resampled to hourly means. Each positive sample comprised the 48 hourly temperature values preceding a recorded positive interval; negative samples were equal-length windows from monitored periods without a recorded estrus event. Windows with five or more missing hours were excluded, whereas retained windows with fewer than five missing values were forward-filled. Although the raw workbook contains step-count readings, the reported model used temperature only. Artificial Intelligence Machine Learning Agriculture Yin, Yapeng Shi, Haiyan Yao, Xuefei Dataset doi:10.17632/kxvsv72sbj.1 oai:data.mendeley.com/kxvsv72sbj.1 Creative Commons Attribution 4.0 International http://creativecommons.org/licenses/by/4.0 https://data.mendeley.com/datasets/kxvsv72sbj 2026-09-01T13:11:40Z