Firefly and Artificial Bee Colony Metaheuristic Algorithms for Enhanced Neural Network Prediction Modeling of Sediment Load in Çoruh River

Katipoğlu O. M.

3rd GLOBAL CONFERENCE on ENGINEERING RESEARCH, 13 - 16 September 2023, pp.500-501

  • Publication Type: Conference Paper / Summary Text
  • Page Numbers: pp.500-501
  • Erzincan Binali Yildirim University Affiliated: Yes


The amount of sediment load (SL) is important for determining the service life of the
downstream dam, river hydraulics, waterworks construction, and reservoir management. The
importance of Artificial Intelligence (AI) and nature-based optimization algorithms is
increasing in solving water resources problems, such as SL estimation. This study combines
the artificial neural network (ANN) algorithm with Firefly (FF) and artificial bee colony
(ABC) optimization techniques for the estimation of monthly SL values in the Çoruh River in
Northeastern Turkey. The estimation of SL values can only be achieved with the inputs of
previous SL and streamflow values, which are provided to the models. The relationship
between actual and predicted sediment values was visually compared using scatter diagrams
and time series comparison charts. The accuracy of the established hybrid approaches was
evaluated according to various statistical metrics such as root means squared error (RMSE),
mean absolute error (MAE), mean absolute percentage error (MAPE), determination
coefficient (R
2), and mean bias error, mean absolute percentage error (MAPE), bias factor
(BF), kling gupta efficiency (KGE). As a result of the analyzes, it was determined that the
ABC-ANN hybrid approach outperformed the others in SL estimation.
As a result of the analyzes, it was determined that the combination of Q(t), Q(t-1) models and
ABC-ANN algorithm with (R
2:0.905, RMSE:1406.730, MAE:769.545, MAPE:5.861, MBE: -
251.090, BF: -4.457, KGE:0.737) had the highest estimation. In addition, since FF and ABC
optimization techniques enable the parameters of the ANN model to be optimized, higher
accuracy prediction outputs were obtained than the single ANN model.