Appraisal of machine learning techniques for predicting emerging disinfection byproducts in small water distribution networks.

Monitoring emerging disinfection byproducts (DBPs) is challenging for many small water distribution networks (SWDNs), and machine learning-based predictive modeling could be an alternative solution. In this study, eleven machine learning techniques, including three multivariate linear regression-based, three regression tree-based, three neural networks-based, and two advanced non-parametric regression techniques, are used to develop models for predicting three emerging DBPs (dichloroacetonitrile, chloropicrin, and trichloropropanone) in SWDNs. Predictors of the models include commonly-measured water quality parameters and two conventional DBP groups. Sampling data of 141 cases were collected from eleven SWDNs in Canada, in which 70 % were randomly selected for model training and the rest were used for validation. The modeling process was reiterated 1000 times for each model. The results show that models developed using advanced regression techniques, including support vector regression and Gaussian process regression, exhibited the best prediction performance. Support vector regression models showed the highest prediction accuracy (R2 =0.94) and stability for predicting dichloroacetonitrile and trichloropropanone, and Gaussian process regression models are optimal for predicting chloropicrin (R2 =0.92). The difference is likely due to the much lower concentrations of chloropicrin than dichloroacetonitrile and trichloropropanone. Advanced non-parametric regression techniques, characterized by a probabilistic nature, were identified as most suitable for developing the predictive models, followed by neural network-based (e.g., generalized regression neural network), regression tree-based (e.g., random forest), and multivariate linear regression-based techniques. This study identifies promising machine learning techniques among many commonly-used alternatives for monitoring emerging DBPs in SWDNs under data constraints.

Evaluation of machine learning techniques to select marine oil spill response methods under small-sized dataset conditions.

Oil spill incidents can significantly impact marine ecosystems in Arctic/subarctic areas. Low biodegradation rate, harsh environments, remoteness, and lack of sufficient response infrastructure make those cold waters more susceptible to the impacts of oil spills. A major challenge in Arctic/subarctic areas is to timely select suitable oil spill response methods (OSRMs), concerning the process complexity and insufficient data for decision analysis. In this study, we used various regression-based machine learning techniques, including artificial neural networks (ANNs), Gaussian process regression (GPR), and support vector regression, to develop decision-support models for OSRM selection. Using a small hypothetical oil spill dataset, the modelling performance was thoroughly compared to find techniques working well under data constraints. The regression-based machine learning models were also compared with integrated and optimized fuzzy decision trees models (OFDTs) previously developed by the authors. OFDTs and GPR outperformed other techniques considering prediction power (> 30 % accuracy enhancement). Also, the use of the Bayesian regularization algorithm enhanced the performance of ANNs by reducing their sensitivity to the size of the training dataset (e.g., 29 % accuracy enhancement compared to an unregularized ANN).

Intelligent computational techniques in marine oil spill management: A critical review.

Effective marine oil spill management (MOSM) is crucial to minimize the catastrophic impacts of oil spills. MOSM is a complex system affected by various factors, such as characteristics of spilled oil and environmental conditions. Oil spill detection, characterization, and monitoring; risk evaluation; response selection and process optimization; and waste management are the key components of MOSM demanding timely decision-making. Applying robust computational techniques based on real-time data (e.g., satellite and aerial observations) and historical records of oil spill incidents may considerably facilitate decision-making processes. Various soft-computing and artificial intelligence-based models and mathematical techniques have been used for the implementation of MOSM's components. This study presents a review of literature published since 2010 on the application of computational techniques in MOSM. A statistical evaluation is performed concerning the temporal distribution of papers, publishers' engagement, research subfields, countries of studies, and selected case studies. Key findings reported in the literature are summarized for two main practices in MOSM: spill detection, characterization, and monitoring; and spill management and response optimization. Potential gaps in applying computational techniques in MOSM have been identified, and a holistic computational-based framework has been suggested for effective MOSM.

Selection of oil spill response method in Arctic offshore waters: A fuzzy decision tree based framework.

A fuzzy decision tree (FDT) based framework was developed to facilitate the selection of suitable oil spill response methods in the Arctic. Hypothetical oil spill cases were developed based on six identified attributes, while the suitability of three spill response methods (mechanical containment and recovery, use of chemical dispersants, and in-situ burning) for each spill case was obtained based on expert judgments. Fuzzy sets were used to address the associated uncertainties, and FDTs were then developed through generating: i) one decision tree for all three response methods (FDT-AP1) and ii) one decision tree for each response method and the development of linear regression models at terminal nodes (FDT-LR). The FDT-LR approach exhibited higher prediction accuracy than the FDT-AP1 approach. A maximum of 100% accurate predictions could be achieved for testing cases using it. On average, 75% of suitable oil spill response methods out of 10,000 performed iterations were predicted correctly.

Segmentation of COVID-19 pneumonia lesions: A deep learning approach.

Background: Lung CT scan has a pivotal role in diagnosis and monitoring of COVID-19 patients, and with growing number of affected individuals, the need for artificial intelligence (AI)-based systems for interpretation of CT images is emerging. In current investigation we introduce a new deep learning-based automatic segmentation model for localization of COVID-19 pulmonary lesions. Methods: A total of 2469 CT scan slices, containing 1402 manually segmented abnormal and 1067 normal slices form 55 COVID-19 patients and 41 healthy individuals, were used to train a deep convolutional neural network (CNN) model based on Detectron2, an open-source modular object detection library. A dataset, including 1224 CT slices of 18 COVID-19 patients and 9 healthy individuals, was used to test the model. Results: The accuracy, sensitivity, and specificity of the trained model in marking a single image slice with COVID-19 lesion were 0.954, 0.928, and 0.961, respectively. Considering a threshold of 0.4% for percentage of lung involvement, the model was capable of diagnosing the patients with COVID-19 pneumonia, with a sensitivity of 0.982% and a specificity of 88.5%. Furthermore, the mean Intersection over Union (IoU) index for the test dataset was 0.865. Conclusion: The deep learning-based automatic segmentation method provides an acceptable accuracy in delineation and localization of COVID-19 lesions, assisting the clinicians and researchers for quantification of abnormal findings in chest CT scans. Moreover, instance segmentation is capable of monitoring longitudinal changes of the lesions, which could be beneficial to patients' follow-up.