Intelligent Water Level Prediction Model for small-and medium-sized rivers Based on Small Sample Data

Abstract

The small-and medium-sized rivers are affected by the perennial dry flow. The historical water level is scarce, which cannot meet the number of training samples in neural network modeling. The model's accuracy cannot meet the requirements when using traditional machine learning methods to predict the water level. An intelligent water level prediction model for small-and medium-sized rivers based on small sample data is proposed to solve this problem. The primary water level prediction model is established by Bayesian linear regression, and the k-nearest neighbor algorithm is introduced to achieve the first correction of the prediction results. The second correction to the water level sequence is performed using the adjacent algorithm to improve the accuracy of the model predictions. Experiments demonstrate that the method can effectively predict water levels in small samples of small-and medium-sized rivers and improve prediction accuracy.