COMPARISON OF STATISTICAL MODEL AND RANDOM FOREST FOR GROUNDWATER CONTAMINATION PATTERNS.
Domain:
environment and energyRecord type:
paper
Creator:
UdeUka
Publisher:
The
Host:
In this study, a random forest model was compared to a statistical model for predicting heavy metal concentrations in groundwater in Edo State, Nigeria. The pH of groundwater samples was determined using a pH meter, and heavy metal concentrations were measured with Atomic Absorption Spectrophotometer (AAS). Pearson Correlation Coefficient was used to evaluate correlations between heavy metal concentrations. Both Random Forest Model (RFM) and Multiple Linear Regression (MLR) were employed to model these concentrations, with goodness of fit assessed via R-squared and root mean square error (RMSE). Results showed that heavy metal concentrations, except for lead, were generally within acceptable limits. The RFM outperformed MLR in predicting iron and lead concentrations but was less effective for arsenic. Python was used for modelling and data extraction. Both models are suitable for predicting groundwater contamination, with RFM showing better overall performance.
Similar
<p>Comparison of random forest model ROC curve.</p>Multivariate Statistical and Machine Learning Assessment of Groundwater Quality to Identify Patterns and Predictors of Nitrate Contamination in Maiduguri Metropolis, NigeriaConfusion matrix of the random forest model.Random Forest model predicted land cover map for Gabon 2015.Random Forest model predicted land cover map for Liberia 2015.Variable importance from the Random Forest model.