Marine Dissolved Oxygen Prediction With Tree Tuned Deep Neural Network

Abstract

Changes in the dissolved oxygen concentration of the ocean have important implications for marine ecosystems and global climate change. However, limited by measurement techniques, the hydrology data is not always complete. Thus, accurate prediction on marine dissolved oxygen concentration (MDOC), is a powerful supplement to the current observation data. Deep neural network is a powerful model to do the prediction, while it is usually difficult and time-consuming to tune its structure. Meanwhile, deep jointly informed neural network (DJINN) provides a user friendly method to tune the structure of neural networks. In this paper, a deep learning-based model called marine deep jointly informed neural network (M-DJINN), is proposed to predict MDOC. M-DJINN improves DJINN performance via initializing the weights of neural network using a zero-mean Gaussian distribution with a variance related to the number of neurons in the neighbor layer. In M-DJINN, to get an ideal deep neural network structure, users only need to tune the tree number and max tree depth. While predicting MDOC on World Ocean Database 2013(WOD13) data with M-DJINN, the novel model proves better than DJINN on both accuracy and convergency. When the max tree depth is set to 10, the mean squared error (MSE) is reduced by 17.6% compared with DJINN. It can be concluded from the experiments that with enough training data, the performance of M-DJINN may keep improving by deepening the neural networks.

Applications of Machine Learning in Chemical and Biological Oceanography

“…However, the combination of different ML algorithms has its advantages, a combination of tree-based models and neural networks termed a Marine-Deep Jointly Informed Neural Network (M-DJINN) estimates the DO in the ocean. 14 The M-DJINN method uses a zero-mean Gaussian distribution to predict the marine DO concentration. By choosing the number of trees and the maximum depth of trees, the M-DJINN proved to be more efficient than DJINN in terms of computing time and prediction ability.…”

“…ML has shown to have better performance in the prediction of DO. However, the combination of different ML algorithms has its advantages, a combination of tree-based models and neural networks termed a Marine-Deep Jointly Informed Neural Network (M-DJINN) estimates the DO in the ocean . The M-DJINN method uses a zero-mean Gaussian distribution to predict the marine DO concentration.…”

Data Reduction in Deterministic Neutron Transport Calculations Using Machine Learning

“…The rectified linear unit (ReLU) was used as the activation function and the Adam optimizer was used to minimize the cost function. While a number of methods have been used to tune NN hyperparmeters [15,16], DJINN has been proven to accurately construct a NN without an extensive optimization search in a number of different settings [17,18]. To ensure the DJINN models used were optimized, the number of trees and the maximum depth of the decision trees were varied and tested on the validation setup.…”

A PCA–EEMD–CNN–Attention–GRU–Encoder–Decoder Accurate Prediction Model for Key Parameters of Seawater Quality in Zhanjiang Bay

“…L. Wang et al used the PSO–SELM–PLS prediction model, and the experimental results showed that the three metrics RMSE, MAPE, and MAE were 0.270, 0.0371, and 0.2284, respectively. L. Wang et al [ 13 ] and J. Liu et al [ 14 ] proposed M-DJINN and wavelet transform-depth Bi–S-SRU prediction models, respectively, and the proposed models achieved 96% prediction accuracy. A. Bilali et al [ 15 ] developed stochastic gradient descent for linear regression (SGD), an artificial neural network (ANN), k-nearest neighbors (k-NN), and support vector machine (SVM) prediction models for chloride concentration and the sodium adsorption ratio using physical parameters as inputs to the models, and SGD and ANN outperformed the other methods in terms of R 2 and RMSE.…”