“…Thus physical models are not feasible for wide employment (Aqil, Kita, Yano, & Nishiyama, 2007). In recent years, machine learning methods have been applied to forecasting long-term streamflow, including artificial neural networks (ANNs; Cheng, Feng, Niu, & Liao, 2015;Esmaeelzadeh, Adib, & Alahdin, 2015;Yu, Qin, Larsen, & Chua, 2014), support vector machines (Zhu, Zhou, Ye, & Meng, 2016), fuzzy algorithms (Dariane & Azimi, 2016;Shi, Hu, Yu, & Hu, 2016), and gray system theory (Ma, Li, Zhang, & Fan, 2013). They perform well for stationary time series, whereas hydrological time series are usually nonstationary processes.…”