Estimation of Incomplete Precipitation Data using the Adaptive Neuro-Fuzzy Inference System (ANFIS) Approach


Katipoğlu O. M.

DATA SCIENCE AND APPLICATIONS, cilt.4, sa.1, ss.11-15, 2021 (Düzenli olarak gerçekleştirilen hakemli kongrenin bildiri kitabı)

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 4 Sayı: 1
  • Basım Tarihi: 2021
  • Dergi Adı: DATA SCIENCE AND APPLICATIONS
  • Sayfa Sayıları: ss.11-15
  • Erzincan Binali Yıldırım Üniversitesi Adresli: Evet

Özet

The completeness and continuity of precipitation data, which is one of the basic components of the hydrological cycle, is of vital importance for the planning of water resources. In this study, the gaps in the missing precipitation data in the Erzincan precipitation observation station were filled by using the adaptive neuro-fuzzy inference system (ANFIS). While Erzincan precipitation station 17094 was used as output, Bayburt 17089, Tercan 17718 and Zara 17716 precipitation stations were selected as model inputs. In the ANFIS model, monthly total precipitation data (52 years) between 1966 and 2017 were used. In the model established, 80% of the data (1968) were used for training and 20% (492) for testing. In the ANFIS model, variables were tried by dividing them into sub-sets between 3 and 8. The most suitable ANFIS model was revealed by comparing various statistical indicators. As a result of the study, 3 sub-sets, hybrid learning algorithm, trimf membership function, and model with 600 epochs were selected as the most suitable model.The completeness a
nd continuity of
precipitation data, which is one of the basic components
of the hydrological cycle, is of vital importance for the
planning of water resources. In this study, the gaps in
the missing precipitation data in the Erzincan
precipitation observa
tion station were filled by using the
adaptive neuro
-
fuzzy inference system (ANFIS). While
Erzincan precipitation station 17094 was used as output,
Bayburt 17089, Tercan 17718 and Zara 17716
precipitation stations were selected as model inputs. In
the ANFI
S model, monthly total precipitation data (52
years) between 1966 and 2017 were used. In the model
established, 80% of the data (1968) were used for
training and 20% (492The completeness a
nd continuity of
precipitation data, which is one of the basic components
of the hydrological cycle, is of vital importance for the
planning of water resources. In this study, the gaps in
the missing precipitation data in the Erzincan
precipitation observa
tion station were filled by using the
adaptive neuro
-
fuzzy inference system (ANFIS). While
Erzincan precipitation station 17094 was used as output,
Bayburt 17089, Tercan 17718 and Zara 17716
precipitation stations were selected as model inputs. In
the ANFI
S model, monthly total precipitation data (52
years) between 1966 and 2017 were used. In the model
established, 80% of the data (1968) were used for
training and 20% (492) for testing. In the ANFIS model,
variables were tried by dividing them into sub
-
sets
between 3 and 8. The most suitable ANFIS model was
revealed by comparing various statistical indicators. As a
result of the study, 3 sub
-
sets, hybrid learning algorithm,
trimf membership function, and model with 600 epochs
were selected as the most suitab
le model.) for testing. In the ANFIS model,
variables were tried by dividing them into sub
-
sets
between 3 and 8. The most suitable ANFIS model was
revealed by comparing various statistical indicators. As a
result of the study, 3 sub
-
sets, hybrid learning algorithm,
trimf membership function, and model with 600 epochs
were selected as the most suitab
le model.