Soil pH Value Forecasting Using UWB Echoes Based on Ensemble Methods

• Main Authors: Chenghao Yang, Jing Liang
• Format: Article
• Language: English
• Published: IEEE 2019-01-01
• Series: IEEE Access
• Subjects: Soil pH values; randomforests; adaboost; gradientboost; bagging predictors
• Online Access: https://ieeexplore.ieee.org/document/8915861/

This paper proposed a new method to predict soil pH values based on ensemble methods via ultra-wideband (UWB) radar echoes, due to the fact that the ensemble method has a fast running speed, fewer parameters, and the amount of data required is not large. 16 categories of UWB soil echoes with different pH values are collected and investigated by 4 types of ensemble methods including bagging, randomforest, adaboost and gradientboost. We use principal component analysis(PCA) to reduce the dimensions of the raw data to reduce the overall amount of computation. First, we applied the PCA algorithm to extract features from the raw signals. Second, we applied these four prediction models to predict the pH values with different feature dimensions. Finally, we compare the prediction performance of these four prediction models with different SNRs. The simulation experiment results show that, when the feature dimension is reduced to 4 to 10, randomforest and bagging provide better performance than adaboost and gradientboost in terms of R² and MSE.