Akademik Veri Yönetim Sistemi
Tezin Türü: Yüksek Lisans
Tezin Yürütüldüğü Kurum: Atatürk Üniversitesi, Fen Bilimleri Enstitüsü, Zootekni Anabilim Dalı, Türkiye
Tezin Onay Tarihi: 2021
Tezin Dili: Türkçe
Öğrenci: ELİF KARTAL
Danışman: Aycan Mutlu Yağanoğlu
Özet:
Objective:
The data of animals in 2 private farms in Erzurum province, Pasinler district
were collected. Milk yield, live weight, milking time, outside temperature,
internal temperature values of these animals were recorded regularly for 90
days. The aim of the thesis is to determine the advantages and disadvantages of
Artificial Neural Networks compared to other models, and it has been determined
which factors affect the most economical milk yield by making more accurate
predictions with this model, and it has been determined that this model will be
used as an alternative.
Method:
Artificial Neural Networks and Multiple Linear Regression Analysis were used
for the estimation of milk yield in the study. By observing the advantages and
disadvantages of each approach, comparisons were made for these two models in
order to select the more suitable model in similar studies. SPSS program was
used for Artificial Neural Networks and Multiple Linear Regression Analysis.
Results:
With 95% R2 value in multiple regression, the explanation power of
milking time, live weight, outside temperature and internal temperature was
found to be high. According to the multiple regression value, the R2 98% value
obtained in artificial neural networks was found to be quite high. In addition,
with the multiple regression analysis, milking time and internal temperature
variables had a significant contribution on milk yield at the selected 5%
significance level (P = 0.000 <0.05). However, the external temperature
independent variable (P = 0.391> 0.05) and the body weight independent
variable (P = 0.353> 0.05) did not significantly contribute to the
regression. In Artificial Neural Networks, it was concluded that milk yield
mostly affects milking time (49.2%), then internal temperature (37%), external
temperature (9%), live weight (4%).
Conclusion:
As a result of the comparison between the two models, it has been determined
that an artificial neural network model can be a more effective and more
effective prediction technique compared to multiple linear regression analysis,
which is a model with high predictive power.