Ice Jam Prediction for Sukhona River Based on KNN and Decision Tree Algorithm
基于KNN和决策树算法的苏霍纳河冰塞预测

摘要

冰塞预测对于寒冷地区减少和预防冰塞洪水具有重要意义。本文主要对冰塞预测在寒冷地区应用的可能性进行评估。Sukhona River基于Russia选定的最重要水文和气象特征，对冰塞进行预测。冰流期间的最高水位和冰塞引起的水位上升是主要决定因素。基于决策树算法的KNN算法开发了最佳预测模型。研究发现，由KNN算法建立的模型表现最佳，并准确地预测了所有堵塞年份。本文的研究为该Veliky Ustyug地区的冰塞预测提供了帮助。knn方法对所1 in研究的河段具有回忆性，比其他预测方法更准确地预测了冰塞的发生。这意味着所选的预测因子具有高度可靠性。

Abstract

Prediction of ice jam is very important for reduction and prevention of ice jam floods in cold regions. This article focuses on the assessment of the possibility of predicting ice jam on the Sukhona River in Russia based on selected most significant hydrological and meteorological features. The maximum water level during the ice drift and ice-jam induced water level rising are the main determinants. The optimal prediction model is developed based on KNN algorithm with decision tree algorithm. The model built by the KNN algorithm was found to perform best and accurately found all blockage years. The research in this paper provides help to establish ice jam prediction in the Veliky Ustyug region. The knn method has a recall of 1 in the studied river segment, which predicts the occurrence of ice jam more accurately than other prediction methods. This implies that the chosen forecast factor is highly reliable.