Repository logo
 

Faculty of Engineering and Built Environment

Permanent URI for this communityhttp://ir-dev.dut.ac.za/handle/10321/9

Browse

Author: search.filters.author.Oyebode, Oluwaseun KunleHas files: Yes

Search Results

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Review of three data- driven modelling techniques for hydrological modelling and forecasting
    (PSP, 2014) Oyebode, Oluwaseun Kunle; Otieno, Fredrick Alfred O.; Adeyemo, Josiah
    Various modelling techniques have been proposed and applied for modelling and forecasting of hydrological sys-tems in different studies. These modelling techniques are majorly categorized into two namely, process-based and data-driven modelling techniques. While the process-based techniques provides detailed description of hydro-logical processes, data-driven techniques however de-scribe the behaviour of hydrological processes by taking into account only limited assumptions about the underly-ing physics of the system being modelled. Although, process-based techniques have been widely applied in numerous hydrological modelling studies, the application of data-driven modelling techniques on the other hand has not been fully embraced in the hydrological domain. This paper provides a comprehensive review of several stud-ies relating to three data-driven modelling techniques namely, K-Nearest Neighbours (K-NN), Model Trees (MTs) and Fuzzy Rule-Based Systems (FRBS). Modern trends with respect to their applications in hydrological model-ling and forecasting studies are also discussed. The struc-ture of this review encapsulates an introduction to each of the modelling techniques, their applications in hydrological modelling and forecasting, identification of areas of con-cern in their use, performance improvement methods, as well as summary of their advantages and disadvantages. The review aims to make a case for the application of data-driven modelling techniques by discussing the benefits em-bedded in its integration into water resources applications.
  • Thumbnail Image
    Item
    Comparison of two data-driven modelling techniques for long-term streamflow prediction using limited datasets
    (SCIELO, 2015-09) Oyebode, Oluwaseun Kunle; Adeyemo, Josiah; Otieno, Fredrick Alfred O.
    This paper presents an investigation into the efficacy of two data-driven modelling techniques in predicting streamflow response to local meteorological variables on a long-term basis and under limited availability of datasets. Genetic programming (GP), an evolutionary algorithm approach and differential evolution (DE)-trained artificial neural networks (ANNs) were applied for flow prediction in the upper uMkhomazi River, South Africa. Historical records of streamflow, rainfall and temperature for a 19-year period (1994-2012) were used for model design, and also in the selection of predictor variables into the input vector space of the model. In both approaches, individual monthly predictive models were developed for each month of the year using a one-year lead time. The performances of the predictive models were evaluated using three standard model evaluation criteria, namely mean absolute percentage error (MAPE), root mean-square error (RMSE) and coefficient of determination (R2). Results showed better predictive performance by the GP models (MAPE: 3.64%; RMSE: 0.52: R2: 0.99) during the validation phase when compared to the ANNs (MAPE: 93.99%; RMSE: 11.17; R2: 0.35). Generally, the GP models were found to be superior to the ANNs, as they showed better performance based on the three evaluation measures, and were found capable of giving a good representation of non-linear hydro-meteorological variations despite the use of minimal datasets.
  • Thumbnail Image
    Item
    Modelling streamflow response to hydro-climatic variables in the Upper Mkomazi River, South Africa
    (2014-06-13) Oyebode, Oluwaseun Kunle; Adeyemo, Josiah; Otieno, Fredrick Alfred O.
    Streamflow modelling remains crucial to decision-making especially when it concerns planning and management of water resources systems in water-stressed regions. This study proposes a suitable method for streamflow modelling irrespective of the limited availability of historical datasets. Two data-driven modelling techniques were applied comparatively so as to achieve this aim. Genetic programming (GP), an evolutionary algorithm approach and a differential evolution (DE)-trained artificial neural network (ANN) were used for streamflow prediction in the upper Mkomazi River, South Africa. Historical records of streamflow and meteorological variables for a 19-year period (1994- 2012) were used for model development and also in the selection of predictor variables into the input vector space of the models. In both approaches, individual monthly predictive models were developed for each month of the year using a 1-year lead time. Two case studies were considered in development of the ANN models. Case study 1 involved the use of correlation analysis in selecting input variables as employed during GP model development, while the DE algorithm was used for training and optimizing the model parameters. However in case study 2, genetic programming was incorporated as a screening tool for determining the dimensionality of the ANN models, while the learning process was further fine-tuned by subjecting the DE algorithm to sensitivity analysis. Altogether, the performance of the three sets of predictive models were evaluated comparatively using three statistical measures namely, Mean Absolute Percent Error (MAPE), Root Mean-Squared Error (RMSE) and coefficient of determination (R2). Results showed better predictive performance by the GP models both during the training and validation phases when compared with the ANNs. Although the ANN models developed in case study 1 gave satisfactory results during the training phase, they were unable to extensively replicate those results during the validation phase. It was found that results from case study 1 were considerably influenced by the problems of overfitting and memorization, which are typical of ANNs when subjected to small amount of datasets. However, results from case study 2 showed great improvement across the three evaluation criteria, as the overfitting and memorization problems were significantly minimized, thus leading to improved accuracy in the predictions of the ANN models. It was concluded that the conjunctive use of the two evolutionary computation methods (GP and DE) can be used to improve the performance of artificial neural networks models, especially when availability of datasets is limited. In addition, the GP models can be deployed as predictive tools for the purpose of planning and management of water resources within the Mkomazi region and KwaZulu-Natal province as a whole.