Search

Large-scale management of surface water resources in urban areas can be difficult, especially if the region is subject to monsoonal waterlogging. Deep learning-based methods for computer vision tasks, such as image segmentation, can effectively be applied to remote sensing data for generating water body maps of large cities, aiding managerial entities, urban planners and policymakers. The robustness of these models to erroneous pixel level training class labels has not been studied in water body segmentation context. Label noise is commonly experienced in classification tasks and may hinder the performance. We collected and densely labeled Sentinel-2 images over Dhaka, one of the most densely populated and flood prone megacities in the world. We synthetically injected four types of label noise viz., i) Gaussian noise, ii) translation, ii) rotation, and iv) mirroring. Our primary objective is to observe and quantitatively analyze the effects of label noise on remote sensing data-driven deep learning models for water body segmentation. Our results show that salt and pepper noise (injected artificially using Gaussian noise) of only 50% can cause a massive 48.55% drop in Intersection-over-Union score. The consequences of learning from training data with different magnitudes and settings of label noise have been explored. [ABSTRACT FROM AUTHOR]

E-commerce platforms have made our life easier and bought plenty of advantages too. However, due to fraud and scams, trust is a concern while buying products online. In this study, we proposed a blockchain-based architecture for the e-commerce sector where data mining technology is used to detect fraudulent users by generating precise and effective rules, and smart contracts are used for enforcement and functionality management within the blockchain network. Our data mining approach yielded a competitive accuracy, precision, recall, and f1-measure of over 99% compared to the state-of-the-art. High performing rules are further tested for ten simulated cycles under completely unseen data to test their rigidity in a real-time scenario. Details analysis of delineation (if any) in rule antecedents has also been analyzed for these cycles. Tested and selected rules are stored in smart contracts deployed within blockchain network for security and immutability. For ensuring strict order criteria maintenance, better return policy, authentic review and scam avoidance, two smart contracts have been written which implement seller reputation mechanism and authentic review maintenance while safeguarding the seller from intentional defaming too. Altogether, the goal of this study is to establish a balance between all the parties within an e-commerce platform so that everyone's right is protected, money is safe and resources are not exploited. [ABSTRACT FROM AUTHOR]

Spatial flood susceptibility mapping (FSM) is one of the key components of flood risk assessment. Recent studies showed the efficacy of hybridized machine learning (ML)-based models in predicting flood susceptibility. The quality of this prediction depends on the presence or absence of label noise in the training data collected from real world flood inundation maps. However, no study has been conducted to explore the effects of label noise in ML based spatial FSM. In the present study, we have explored the effects of four different random class label noises on the regression performances and model complexities of both standalone and hybridized ML based FSM. We also explored the optimal hyperparameter values of different hybridized ML and reported the robustness of each model in the presence and absence of label noise. The hybridized ML models showed greater efficacy in modeling spatial flood susceptibility in both cases. [ABSTRACT FROM AUTHOR]