Explainable machine learning for the prediction and assessment of complex drought impacts.

Drought is a common and costly natural disaster with broad social, economic, and environmental impacts. Machine learning (ML) has been widely applied in scientific research because of its outstanding performance on predictive tasks. However, for practical applications like disaster monitoring and assessment, the cost of the models failure, especially false negative predictions, might significantly affect society. Stakeholders are not satisfied with or do not "trust" the predictions from a so-called black box. The explainability of ML models becomes progressively crucial in studying drought and its impacts. In this work, we propose an explainable ML pipeline using the XGBoost model and SHAP model based on a comprehensive database of drought impacts in the U.S. The XGBoost models significantly outperformed the baseline models in predicting the occurrence of multi-dimensional drought impacts derived from the text-based Drought Impact Reporter, attaining an average F2 score of 0.883 at the national level and 0.942 at the state level. The interpretation of the models at the state scale indicates that the Standardized Precipitation Index (SPI) and Standardized Temperature Index (STI) contribute significantly to predicting multi-dimensional drought impacts. The time scalar, importance, and relationships of the SPI and STI vary depending on the types of drought impacts and locations. The patterns between the SPI variables and drought impacts indicated by the SHAP values reveal an expected relationship in which negative SPI values positively contribute to complex drought impacts. The explainability based on the SPI variables improves the trustworthiness of the XGBoost models. Overall, this study reveals promising results in accurately predicting complex drought impacts and rendering the relationships between the impacts and indicators more interpretable. This study also reveals the potential of utilizing explainable ML for the general social good to help stakeholders better understand the multi-dimensional drought impacts at the regional level and motivate appropriate responses.

Feature selection approaches for predictive modelling of cadmium sources and pollution levels in water springs.

The World Health Organization lists cadmium (Cd) as one of the top ten chemicals of public health concern. Cd is toxic at relatively low exposure levels and has acute and chronic effects on both health and the environment. In this study, we investigate a suite of data-driven methods that could assist decision-makers in estimating Cd levels in water springs, and in identifying polluting sources. Machine learning (ML) regression models were used to identify sources of contamination and predict Cd levels based on support vector machines and a variety of tree-based models, including Random Forests, M5Tree, CatBoost, and gradient boosting. Feature selection analysis revealed that heavy traffic and distance to a major power plant in the sampled area play a leading role in springs Cd contamination, together with precipitation levels and average of slopes of the closest waste dumps upstream to sampled springs. Our best performing ML model was the Adaboost regression tree using all the features (RMSE = 19.36, R^2 = 0.64). Our findings highlight the effectiveness of predictive data-driven modeling in addressing environmental challenges, particularly in high-risk areas with low resources.

Meta-learning for fake news detection surrounding the Syrian war.

In this article, we pursue the automatic detection of fake news reporting on the Syrian war using machine learning and meta-learning. The proposed approach is based on a suite of features that include a given article's linguistic style; its level of subjectivity, sensationalism, and sectarianism; the strength of its attribution; and its consistency with other news articles from the same "media camp". To train our models, we use FA-KES, a fake news dataset about the Syrian war. A suite of basic machine learning models is explored, as well as the model-agnostic meta-learning algorithm (MAML) suitable for few-shot learning, using datasets of a modest size. Feature-importance analysis confirms that the collected features specific to the Syrian war are indeed very important predictors for the output label. The meta-learning model achieves the best performance, improving upon the baseline approaches that are trained exclusively on text features in FA-KES.

Prenatal exposure to criteria air pollutants and associations with congenital anomalies: A Lebanese national study.

Maternal exposure to air pollution has been associated with a higher birth defect (BD) risk. Previous studies suffer from inaccurate exposure assessment methods, confounding individual-level variations, and classical analytical modelling. This study aimed to examine the association between maternal exposure to criteria air pollutants and BD risk. A total of 553 cases and 10,214 controls were identified from private and public databases. Two subgroups were then formed: one for a matched case-control design, and another for Feature Selection (FS) analysis. Exposure assessment was based on the mean air pollutant-specific levels in the mother's residential area during the specific BD gestational time window of risk (GTWR) and other time intervals. Multivariate regression models outcomes consistently showed a significant protective effect for folic acid intake and highlighted parental consanguinity as a strong BD risk factor. After adjusting for these putative risk factors and other covariates, results show that maternal exposure to PM2.5 during the first trimester is significantly associated with a higher overall BD risk (OR:1.05, 95%CI:1.01-1.09), and with a higher risk of genitourinary defects (GUD) (OR:1.06, 95%CI:1.01-1.11) and neural tube defects (NTD) (OR:1.10, 95%CI:1.03-1.17) during specific GTWRs. Maternal exposure to NO2 during GTWR exhibited a significant protective effect for NTD (OR:0.94, 95%CI:0.90-0.99), while all other examined associations were not statistically significant. Additionally, maternal exposure to SO2 during GTWR showed a significant association with a higher GUD risk (OR:1.17, 95%CI:1.08-1.26). When limiting selection to designated monitor coverage radiuses, PM2.5 maintained significance with BD risk and showed a significant gene-environment interaction for GUD (p = 0.018), while NO2 protective effect expanded to other subtypes. On the other hand, FS analysis confirmed maternal exposure to PM2.5 and NO2 as important features for GUD, CHD, and NTD. Our findings, set the basis for building a novel BD risk prediction model.