Classifying Milk Yield Using Deep Neural Network

Creative Commons License


PAKISTAN JOURNAL OF ZOOLOGY, vol.52, no.4, pp.1319-1325, 2020 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 52 Issue: 4
  • Publication Date: 2020
  • Doi Number: 10.17582/journal.pjz/20190527090506
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Agricultural & Environmental Science Database, Animal Behavior Abstracts, Aquatic Science & Fisheries Abstracts (ASFA), BIOSIS, CAB Abstracts, Environment Index, Geobase, Veterinary Science Database
  • Page Numbers: pp.1319-1325
  • Keywords: Deep neural network, Milk yield, Lactation, First calving, Classification, LACTATION NUMBER, DAIRY, AGE, TRAITS, REPRODUCTION, HOLSTEINS, STAGE, LIFE
  • Akdeniz University Affiliated: Yes


This study aim to describe the impact of the number of lactation, lactation days, age at first calving and breeding, and number of insemination (ratio) on cattle milk yield (last seven days in average). For this purpose, the milk yields of 156 Holstein Friesian cattle were investigated according to different age, lactation, calving and insemination associated parameters. Optimum values in literature were organized by an expert in establishing classification data. The expert determined the classes of the outputs data (average milk) through the input data (calving age, milking days, number of lactation and insemination). Applying deep neural networks, we established that average classification success of the system was 69.23% as a result of 6-Layers Cross-Verification Test which is commonly used in the literature for small datasets. In these datasets, it was found that 84 animals had GOOD, 39 animals carried POOR and 33 animals possessed MEDIUM milk yield. It was revealed that there is provided animal raising conditions by 53,84% (84/156*100); therefore, there is no professional farm management. Taken together, the finding show that there is a need of additional controlled management on animal raising and mistakes of the enterprise need to be recovered as early as possible.