Non-Linear Input Variable Selection Approach Integrated With Non-Tuned Data Intelligence Model for Streamflow Pattern Simulation

Abstract

Streamflow modeling is considered as an essential component for water resources planning and management. There are numerous challenges related to streamflow prediction that are facing water resources engineers. These challenges due to the complex processes associated with several natural variables such as non-stationarity, non-linearity, and randomness. In this study, a new model is proposed to predict long-term streamflow. Several lags that cover several years are abstracted using the potential of Extreme Gradient Boosting (XGB) then after the selected inputs variables are imposed into the predictive model (i.e., Extreme Learning Machine (ELM)). The proposed model is compared with the stand-alone schema in which the optimum lags of the variables are supplied into the XGB and ELM models. Hydrological variables including rainfall, temperature and evapotranspiration are used to build the model and predict the streamflow at Goksu-Himmeti basin in Turkey. The results showed that XGB model performed an excellent result in which can be used for predicting the streamflow pattern. Also, it is clear from the attained results that the accuracy of the streamflow prediction using XGB technique could be improved when the high number of lags was used. However, the implementation of the XGB is tree-based technique in which several issues could be raised such as overfitting problem. The proposed schema XGBELM in which XGB approach is selected the correlated inputs and ranking them according to their importance; then after, the selected inputs are supplied into the ELM model for the prediction process. The XGBELM model outperformed the stand-alone schema of both XGB and ELM models and the high-lagged schema of the XGB. It is important to indicate that the XGBELM model found to improve the prediction ability with minimum variables number.INDEX TERMS Correlated variables, non-linear XGB approach, extreme learning machine, streamflow simulation.

Streamflow Prediction Using Deep Learning Neural Network: Case Study of Yangtze River

“…With Non-Tuned Data Intelligence Model was introduced by Hadi [34] to forecast the streamflow. Several hydrological variables including rainfalls, temperature and evapotranspiration were used to build and train the model.…”

Hybrid Machine Learning Ensemble Techniques for Modeling Dissolved Oxygen Concentration

“…Although the AI models proved robust, the determination of appropriate input variables is the major problem in most of the techniques. According to Hadi et al [49] and Yaseen et al [53], excessive inputs deteriorate the model performance while too little inputs may not reveal all of the hidden information in the time series. Therefore, in this study, nonlinear sensitivity analysis was carried out using MLP owing to its promising capability for modeling DO concentration and other WQ parameters.…”

“…Except for TS, all the WQ variables show excellent inverse relationship with the DO. Even though studies such as [45], [48], [49], [61], [81] have criticized the classical linear input variable selection and recommended the use of nonlinear approaches; they are applicable for input selection and the determination of linear patterns between the variables. The observed WQ parameters were analysed, and the statistical overview of the data was obtained as presented in Table 1.…”

“…One of the factors affecting the accuracy of the models is the model input determination, which depends on the identification of the AI-based models, and others include model configuration, prediction horizon, etc. According to Hadi et al [50] different approaches have been reported for determining the most suitable input variables including principle component analysis (PCA), autocorrelation function (ACF), partial autocorrelation function (PACF) and Pearson correlation analysis. Conversely, those techniques are associated to the input-output linear relationship [51].…”

“…The selection of 4 input combinations is predominantly based on the nonlinear sensitivity analysis. The advantage of using the nonlinear sensitivity input variables selection approach to carefully determine the most relevant factors has been reported in various studies [49], [78], [79]. Therefore, a sensitivity analysis using MLP between the observed WQ parameters was evaluated, and the average values of RMSE and NSE were considered and ranked in hierarchical order of the WQ variables as presented in Table 2.…”