Improving the accuracy of K-nearest neighbour method in long-lead hydrological forecasting

The nonparametric regression method of K-Nearest Neighbour (K-NN) has been used in a variety of eco-hydrological issues. In this study, some techniques were presented to improve the accuracy of the K-NN method in forecasting accumulated 9- month in ow, from 1971 to 2001, of Zayandeh-rud dam in Iran, from winter to the end of the following summer. The considered improving techniques consisted of: 1) selecting the best data preprocessing functions, 2) selecting the best number of neighbours, 3) selecting the best distance functions, 4) specifying the best weights of predictors at distance functions, and 5) adding the ability of extrapolation to K-NN using a proposed method. Final results showed that the use of the mentioned techniques had promoted the accuracy of K-NNʹs forecast, meaningfully. The results of goodness-of-t criteria for the optimized K-NN in comparison with a regular K-NN presented an increase by 31% in correlation coecient (from 65% to 96%), a decrease from 31% to 8% in volume error, and nally a drop from 54% to 25% in the root mean square error.