Combined model water level prediction method based on similarity search

Abstract

The invention discloses a combined model water level prediction method based on similarity search. Water level of previous days related to a day to be predicted is confirmed to be sequences to be matched by utilizing correlation coefficient. A series of water level time sequences which are not similar to the sequences to be matched are searched from historical data on the basis of similarity search and eliminated from original time sequences and then act as training sets of a prediction model. The method mainly comprises data preprocessing, similarity search and a combined prediction model. Data preprocessing aims at filling gap data and restoring error data. According to similarity search, a series of time sequences which are not similar to the sequences to be matched are eliminated from the historical data of previous years by utilizing dynamic bending distance and the fixed slide window technology. The combined prediction model has two basic models: a BP neural network improved by an LM algorithm and a support vector machine, and proportion of the basic models in current prediction is dynamically adjusted by utilizing the Bayes theorem according to prediction performance of each basic model at the previous moment. The high-precision and real-time requirements required by flood prevention and disaster resistance can be realized.

Description

technical field [0001] The invention relates to a method for predicting water level based on a combined model of similarity search, in particular to the preprocessing of hydrological water level data, the use of similarity search to effectively reduce the dimension of the training set, and the improved BP neural network and support vector based on LM algorithm The invention discloses a machine combination forecasting method, which belongs to the field of information technology. Background technique [0002] With the advancement of the times and the wide application of computers, more and more content is stored on the computer. How to mine useful information from these massive historical data to serve the prediction of the future, and at the same time, it will not be lost in the historical data. Redundancy, vacancies and wrong information interference have become a topic of concern to people. Especially in the field of hydrology, China has a large number of hydrological surv...

AI Technical Summary

Problems solved by technology

The most widely used are hydrological prediction models established for specific watersheds, such as the Xin'anjiang model, etc. These models have a certain scope of application and can only be mastered by some professionals, so they are not very adaptable

Images