A Hybrid ANFIS-GA Approach for Estimation of Regional Rainfall Amount


CALP M. H.

GAZI UNIVERSITY JOURNAL OF SCIENCE, vol.32, no.1, pp.145-162, 2019 (ESCI) identifier identifier

  • Publication Type: Article / Article
  • Volume: 32 Issue: 1
  • Publication Date: 2019
  • Journal Name: GAZI UNIVERSITY JOURNAL OF SCIENCE
  • Journal Indexes: Emerging Sources Citation Index (ESCI), Scopus, TR DİZİN (ULAKBİM)
  • Page Numbers: pp.145-162
  • Keywords: Rainfall Amount, Estimation, Artificial Intelligence, ANFIS, Genetic Algorithm, ARTIFICIAL NEURAL-NETWORKS, ALGORITHM, SYSTEMS, MODEL, IDENTIFICATION, OPTIMIZATION, SIMULATION, PREDICTION
  • Karadeniz Technical University Affiliated: Yes

Abstract

Effective use and management of ever-diminishing water resources are critically important to the future of humanity. At this point, rainfall is one of the most important factors that supply water resources, but the fact that the rainfall higher is more than normal causes many disasters such as flood, erosion. Therefore, rainfall amount must be analyzed mathematically, statistically or heuristically in order to take precautions, in the region. In this study, an Adaptive Neuro Fuzzy Inference System - Genetic Algorithm (ANFIS-GA) based hybrid model was proposed for estimation of regional rainfall amount. Purpose of the study is to minimize the loss of life and goods for people of the region by estimating the amount of annual rainfall and ensuring effective management of water resources and allowing some evaluations and preparations according to possible climate changes. The estimation model was developed by coding in the MATLAB package program. In the development of the model, 3650 meteorological data from 2008-2018 years belonging to Basel, a Swiss city, were utilized. The real data were tested on both the Artificial Neural Network (ANN) and the hybrid ANFIS-GA model. The obtained results demonstrated that the training R-value of the suggested ANFIS-GA model was 0.9920, the testing R-value was 0.9840 and the error ratio was 0.0011. This clearly shows that predictive performance of the model is high and error level is low, and therefore that hybrid approaches such as ANFIS-GA can be easily used in predicting meteorological events.