Technical note: A dissolved oxygen prediction method based on K-means clustering and the ELM neural network: A case study of the Changdang Lake, China

According to mass data from the internet, prediction is an important application of information processing. Predicting the amount of dissolved oxygen (DO) of a water environment is vital and can improve the aquaculture production efficiency. Using water quality and meteorological data for the Changdang Lake (Jiangsu, China) from 2014 to 2016, this study proposes a new strategy for predicting short-term DO based on K-means clustering and extreme learning machine (ELM) neural networks. First, the weights of the environmental factors on the DO are obtained using Pearson correlation analysis, and similar days are defined. Subsequently, according to the defined similarity, the K-means clustering algorithm divides the historical data into several clusters and identifies similar sample sets that have characteristics that are similar to the forecast day. Finally, after the ELM neural network model establishes the training and testing data, the DO is predicted using the similar sample set and the real-time environmental factors of the forecast days as input data. The prediction efficiency of our model was compared to that of others in terms of the mean absolute percentage error and the mean square error. The experimental results showed that our approach had higher forecasting accuracy and faster computation speed, which is beneficial for water management.