Article
Ecology
Lalit Kumar, Mohammad Saud Afzal, Ashad Ahmad
Summary: The measurement of marine water quality is a crucial research topic for environmental and ocean modelers. This study aims to predict turbidity in the marine environment of Hong Kong using machine learning approaches. Different machine learning models were compared, and the results show that the LSTM-RNN model outperformed the others with an accuracy of 88.45%.
REGIONAL STUDIES IN MARINE SCIENCE
(2022)
Article
Engineering, Civil
Nick Hall, Ashley Rust, Terri S. Hogue, Kamini Singha
Summary: This study explores the characteristics of watersheds that experience increased turbidity after the 2013 West Fork Complex Fire in Colorado. The findings show that burned watersheds have higher turbidity spikes following precipitation events compared to unburned watersheds. The severity of burn and vegetation recovery in the watersheds also influence the correlation between total storm volume and turbidity.
JOURNAL OF HYDROLOGY
(2022)
Article
Chemistry, Analytical
Chen-Hua Chu, Yu-Xuan Lin, Chun-Kuo Liu, Mei-Chun Lai
Summary: Given progress in water-quality analytical technology and the emergence of the Internet of Things (IoT) in recent years, compact and durable automated water-quality monitoring devices with dual light sources have been developed. These devices can measure scattering, transmission, and reference light simultaneously, making them suitable for monitoring water quality in low and high turbidity conditions.
Article
Environmental Sciences
Jungsu Park, Jin Chul Joo, Ilsuk Kang, Woo Hyoung Lee
Summary: This study employed two ensemble learning models, XGBoost and LGB, to predict turbidity (T) in water, which is crucial for effective water quality management. The input variables were classified into three groups based on the flow phase, and different time-frequency datasets were utilized to develop the models. The results showed that the model (Model 1) using data classified into three phases outperformed the model without classification (Model 2). Further analysis revealed that the performance differences between Models 1 and 2 were determined by the different data distributions in the three phases. By considering these differences, Model 1 exhibited better performance compared to Model 2. The Shapley additive explanation (SHAP) provided a reasonable interpretation of the difference in model predictions between Models 1 and 2.
ENVIRONMENTAL EARTH SCIENCES
(2023)
Article
Water Resources
Bhargav Rele, Caleb Hogan, Sevvandi Kandanaarachchi, Catherine Leigh
Summary: Many water-quality monitoring programs cannot afford to distribute turbidity sensors throughout networks. In this study, we compared the performance of different models (ARIMA, LSTM, and GAM) in forecasting stream turbidity using low-cost in-situ sensors and publicly available databases. We found that ARIMA and GAM provided the most accurate predictions, and we also constructed a meta-model that outperformed all other models. The study suggests that this methodology can achieve high accuracy in predicting turbidity, especially when cost prohibits the use of direct turbidity sensors.
HYDROLOGICAL PROCESSES
(2023)
Article
Environmental Sciences
Ricardo Juncosa, Jose Luis Cereijo, Ricardo Vazquez
Summary: Water is essential for human life, and issues like turbidity and color changes in water supply distribution systems are common in many cities. Investigating the processes and variables that promote these episodes is necessary, with a focus on sedimentation and resuspension of particles in the system.
Article
Agriculture, Multidisciplinary
Hannah Wenng, Robert Barneveld, Marianne Bechmann, Hannu Marttila, Tore Krogstad, Eva Skarbovik
Summary: The study aimed to identify dominant sediment runoff processes in cultivated grain-dominated catchments in a cold climate. Assessment of turbidity data, catchment properties, and agricultural management data revealed a clockwise concentration-discharge hysteresis pattern in both catchments, with discharge being the main driver for turbidity. Soil tillage intensity and index of connectivity also impacted the hysteresis index.
AGRICULTURE ECOSYSTEMS & ENVIRONMENT
(2021)
Article
Environmental Sciences
Cecile Vuilleumier, Pierre-Yves Jeannin, Marc Hessenauer, Pierre Perrochet
Summary: This study discusses the sources and transport pathways of turbidity in karst springs, revealing the mechanisms of turbidity changes during flood events through observations and numerical simulations. Turbidity mainly comes from underground sediment and soil, and is related to the average boundary shear stress within the conduit network.
WATER RESOURCES RESEARCH
(2021)
Article
Environmental Sciences
Frederic Anderson Konkobo, Paul Windinpsidi Savadogo, Mamounata Diao, Roger Dakuyo, Mamoudou Hama Dicko
Summary: This study explores the potential of plant extracts as biocoagulants in raw water treatment, and finds that seeds of Acacia nilotica, Adansonia digitata, Balanites aegyptiaca, Tamarindus indica, Aloe vera sap, and Opuntia ficus indica sap show varying levels of coagulant activity. Moringa oleifera seeds and Boscia senegelensis seeds are found to be highly effective alternatives to aluminum sulfate, reducing turbidity and pathogenic microorganisms in water.
FRONTIERS IN ENVIRONMENTAL SCIENCE
(2023)
Article
Agronomy
Hyojin Kwon, Eunsom Choe, Md. Iqbal Hossain, Ki-Hwan Park, Dong-Un Lee, Changsun Choi
Summary: This study investigated the correlation between water quality and microbial levels during the processing of salted Chinese cabbages. The results showed that the wash water and cabbage samples were contaminated with bacteria but not Escherichia coli. There was a high positive correlation between turbidity and bacterial levels, suggesting that real-time monitoring of turbidity can help maintain the cleaning effect.
POSTHARVEST BIOLOGY AND TECHNOLOGY
(2023)
Article
Environmental Sciences
Hamdhani Hamdhani, Drew E. Eppehimer, David Walker, Michael T. Bogan
Summary: The handheld fluorometer showed good performance in measuring chlorophyll-a, being insensitive to ambient light and turbidity. It performed well at low chlorophyll-a concentrations and moderate turbidity levels, but had lower accuracy with high chlorophyll-a concentrations and low turbidity levels. A calibration equation was developed to improve accuracy in the latter conditions for field monitoring of potential harmful algal blooms.
Article
Environmental Sciences
Wenxiang Zhang, Dan Zhang, Benwei Shi, Zhonghao Zhao, Jianxiong Sun, Yujue Wang, Xing Wang, Yang Lv, Yue Li, Youcai Liu
Summary: Turbidity is an important parameter in monitoring water quality, and laboratory experiments analyzing the effect of algal concentration on turbidity measurement can improve its accuracy. The results of this study indicate that algal concentration significantly affects measured turbidity, particularly at low turbidity levels.
Article
Environmental Sciences
Sharif Hossain, Guna A. Hewa, Christopher W. K. Chow, David Cook
Summary: This study compares two different approaches to model monochloramine decay in a water distribution system. The results suggest that the data analytics model has relatively higher accuracy in predicting monochloramine residual concentrations.
Article
Environmental Sciences
Michael S. Wetz, Nicole C. Powers, Jeffrey W. Turner, Yuxia Huang
Summary: During the COVID-19 quarantine period, some highly impacted coastal regions saw a localized reduction in fecal indicator bacteria, while less impacted regions showed no widespread improvement in water quality. The study emphasizes the ephemeral nature of coastal water quality improvements during the quarantine period, which are mainly reserved for the most severely affected systems.
SCIENCE OF THE TOTAL ENVIRONMENT
(2022)
Article
Environmental Sciences
Marcelle Lock, Neil Saintilan, Iris van Duren, Andrew Skidmore
Summary: The Australian New South Wales Estuary health assessment and biodiversity monitoring program has set state-wide targets for estuary health. In this study, the use of remote sensing derived data for monitoring water quality indicators in selected lakes along the coast was investigated. The results showed that the remote sensing products were partly successful in predicting chlorophyll a concentration and water clarity, but varied across years and lakes. It is likely that the physical differences between the systems influence the algorithm's output, suggesting the need for a tailored monitoring approach.
Article
Management
Matthew Stevenson, Christophe Mues, Cristian Bravo
Summary: Compared to consumer lending, mSME credit risk modeling is more challenging due to limited data availability, with textual loan assessment being a standard practice. Deep Learning and NLP techniques, including the BERT model, are used to extract information from textual assessments, showing surprisingly effective prediction of default. However, combining text with traditional data does not enhance predictive capability, with performance varying based on text length. Our proposed Deep Learning model is robust to text quality and can partly automate the mSME lending process.
EUROPEAN JOURNAL OF OPERATIONAL RESEARCH
(2021)
Article
Computer Science, Artificial Intelligence
Luisa Roa, Alejandro Correa-Bahnsen, Gabriel Suarez, Fernando Cortes-Tejada, Maria A. Luque, Cristian Bravo
Summary: This paper investigates the impact of alternative data from an app-based marketplace on credit scoring models, revealing that these new data sources are particularly effective for predicting financial behavior in low-wealth and young individuals. Additionally, the study shows interesting non-linear trends in the variables from the app, which are normally invisible to traditional banks.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Management
Maria Oskarsdottir, Cristian Bravo
Summary: This study presents a multilayer network model for credit risk assessment and finds that including centrality multilayer network information in the model can significantly improve predictive gains. The results suggest that default risk is highest when an individual is connected to many defaulters, but this risk can be mitigated by the size of the individual's neighborhood, showing that default risk and financial stability propagate throughout the network.
OMEGA-INTERNATIONAL JOURNAL OF MANAGEMENT SCIENCE
(2021)
Article
Geography, Physical
Matthew Stevenson, Christophe Mues, Cristian Bravo
Summary: LiDAR technology provides detailed three-dimensional elevation maps of urban and rural landscapes. This paper proposes a convenient task-agnostic tile elevation embedding method using unsupervised Deep Learning. The potential of the embeddings is tested by predicting deprivation indices, showing improved performance compared to using standard demographic features alone. The paper also demonstrates the coherent tile segments generated by the embedding pipeline using Deep Learning and K-means clustering.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2022)
Article
Public, Environmental & Occupational Health
Mario P. Brito, Matthew Stevenson, Cristian Bravo
Summary: This study explores the use of machine learning methods in simulating expert risk assessments, proposes a natural language-based probabilistic risk assessment model, and validates its feasibility.
Article
Management
Kamesh Korangi, Christophe Mues, Cristian Bravo
Summary: In this paper, the study focuses on analyzing mid-cap companies using a large dataset of US mid-cap companies observed over 30 years. The researchers use transformer models to predict default probability term structure and determine the most influential data sources for default risk. The results show that the proposed deep learning architecture outperforms traditional models and provides an importance ranking for different data sources using a Shapley approach.
EUROPEAN JOURNAL OF OPERATIONAL RESEARCH
(2023)
Article
Computer Science, Artificial Intelligence
Ricardo Munoz-Cancino, Cristian Bravo, Sebastian A. Rios, Manuel Grana
Summary: This paper introduces an information-processing framework that combines feature engineering, graph embeddings, and graph neural networks to improve credit scoring models. The results show that this approach outperforms traditional methods in assessing creditworthiness using social interaction data. Additionally, in the field of corporate lending, considering the relationships between companies and other entities is crucial for evaluating thin-file companies. The study also highlights the significant value of graph data in helping companies with little or no credit history enter the financial system.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Ricardo Munoz-Cancino, Cristian Bravo, Sebastian A. Rios, Manuel Grana
Summary: Credit risk management has been using credit scoring models at different stages for over half a century. Social network data has been shown to increase the predictive power of these models, especially when historical data is limited. This study analyzes the dynamics of creditworthiness assessment and finds that credit scoring based on borrowers' history improves performance initially and then stabilizes. The use of social network features adds value to credit scoring for loan applications and throughout the study period for business scoring.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Proceedings Paper
Computer Science, Interdisciplinary Applications
Pierre Miasnikof, Alexander Y. Shestopaloff, Cristian Bravo, Yuri Lawryshyn
Summary: Graph isomorphism is an intractable problem, and computing graph similarity metrics is NP-hard. However, assessing (dis)similarity between networks is crucial in various fields. This article proposes a statistical approach to answer questions about network similarity and difference, using distance matrices and probability distributions. The comparison focuses on connectivity and community structure rather than observable graph characteristics. Experimental results with synthetic and real-world graphs validate the effectiveness and accuracy of the proposed technique.
COMPLEX NETWORKS AND THEIR APPLICATIONS XI, COMPLEX NETWORKS 2022, VOL 2
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Ricardo Munoz-Cancino, Cristian Bravo, Sebastian A. Rios, Manuel Grana
Summary: Credit scoring models are the primary instrument used by financial institutions to manage credit risk. However, research on behavioral scoring is scarce due to difficulties in data access. This study presents a methodology for evaluating model performance when trained with synthetic data and applied to real-world data. Results show that the quality of synthetic data decreases as the number of attributes increases, and models trained with synthetic data show a slight reduction in performance compared to those trained with real data.
HYBRID ARTIFICIAL INTELLIGENT SYSTEMS, HAIS 2022
(2022)
Article
Business, Finance
Daniela Lazo, Raffaella Calabrese, Cristian Bravo
JOURNAL OF CREDIT RISK
(2020)
Article
Computer Science, Artificial Intelligence
David Barrera Ferro, Sally Brailsford, Cristian Bravo, Honora Smith
DECISION SUPPORT SYSTEMS
(2020)