Chandrakona Vidyasagar Mahavidyalaya

Commercial financial institutions lose money from non-performing debts, as classically credit scoring models are based on static structured features and do not account for behavioural signals and early warning triggers, and struggle to preserve privacy and fairness across multiple branches. This work proposes a hybrid deep learning architecture with LSTM and Transformer encoder using collaborative filtering that allows to model 6-month behavioural history and latent pattern extraction in a federated averaging for privacy preserving training without centralising the sensitive data. A dynamic watchlist trigger with a composite loss of Binary Cross Entropy, L1 Sparsity and fairness regularisation for early detection of distress, interpretability and Equal Opportunity Difference. Experiments on a commercial bank dataset with 548,000 records of 38,000 clients yields an AUC of 0.919 and F1-score of 0.234 and improved risk detection and return rate analysis as compared to the baseline models and reduced Equal Opportunity Difference.

The peripheral villages of the Indian Sundarban demonstrate considerable vulnerability to the impacts of climate change. This study assesses present socioeconomic vulnerability and resilience of rural livelihoods to climate change at the local level. Alongside verifies these parameters with an Artificial Neural Network (ANN) prediction model in selected villages in the Gosaba and Hingalganj CD Block, located on the fringes of the Sundarban Mangrove Forest (SMF).This study also aims to assess socio-economic vulnerability and climate resilience nexus. The current study employed a comprehensive vulnerability framework, evaluating exposure, adaptive capacity, and sensitivity index by weighting the initial eigenvalues of each indicator based on their variance percentage via Principal components analysis (PCA). Household resilience index (HRI) was determined by assessing the Social and Physical Resilience (SPR), Environmental Resilience(ER), Socioeconomic Resilience (SER).Selected villages of Gosaba CD Block are relatively highly vulnerable, however the performance of HRI is comparatively better. Sampled villages of Hingalganj CD Block demonstrate moderate vulnerability while possessing relatively low to moderate resilience.In order to mobilize and actively engage local communities, it will be crucial to develop Climate Resilient Villages (CRVs) and establish institutional mechanisms at the village level, such as farmers’ cooperative societies, self-help groups, climate risk management committees.In formulating national disaster management, social welfare, or resource management strategies, it is imperative to devise specific action plans for the designated localities. The national government must facilitate efficient decentralization of governance, allowing local governments to achieve certain objectives.

In this study a model for emergency medical services(EMS) in a smart city is proposed to enhance the EMS efficiency by coordinating IoT, Bluetooth technology, geographic information system (GIS), and the scheduling strategy of traveling salesman problem (TSP). The system collects real-time patient data from IoT-enabled medical devices attached with the beneficiaries and uses GIS to optimize the schedule dynamically for the emergency vehicle. The system dynamically collects and analyzes data on patient’s vitals to identify the houses needing medical assistance and the nature of assistance. Then medical assistance team is sent by selecting shortest route through the selected houses to provide medical assistance. Traffic congestion in the route is monitored using GIS facility and the selected route is modified accordingly. Route selection through the selected locations (houses) can be treated as a TSP, where distance between any two locations can be found from GIS. Hence, an efficient and consistent algorithm for the TSPs is suggested and is used to develop a smart EMS with the help of IoT and GIS. in the algorithm, at first a procedure is used to generate a set of potential solutions(Hamiltonian paths through the target houses). Then, another procedure is used to explore the search space properly with the help of some predefined perturbation rules. If a selected rule (randomly selected) for the perturbation of a solution fails to improve the same then K-opt is used for possible enhancement. Another procedure is used for the regeneration of the stagnant solutions to overcome any local optima. The second and third procedures are repeated iteratively for searching the best schedule. The testing of the approach is done using different test instances from the TSPLIB and its efficiency and accuracy for considerably large size TSPs is well established. Using this heuristic, a case study in an urban setting is done to demonstrate the effectiveness of the EMS. The integration of IoT, GIS, and the proposed heuristic for the TSPs not only reduces response time but also enhances overall EMS efficiency, suggesting a promising solution for urban health-care systems aiming to improve emergency response and public health outcomes.

This study presents a model for emergency medical assistance services (EMAS) in a smart city efficiently by integrating Internet of Things (IoT) technology with the scheduling approach travelling salesman problem (TSP). The system utilizes real-time data from IoT-enabled medical devices attached with the beneficiaries and GIS (geographic information system) to optimize the schedule of service vehicle dynamically. The system continuously collects and processes data on patient vitals to identify the houses needing medical assistance and the nature of assistance. Then medical assistance team is sent by selecting shortest route through the selected houses to provide medical assistance. Traffic congestion in the route is monitored using GIS facility and the selected route is modified accordingly. Route selection through the selected locations (houses) can be treated as a TSP, where, the distance between any two locations can be found from GIS. So an efficient algorithm for the TSPs with significant high accuracy is required for the same. In this study such an algorithm for the TSPs is suggested and is used to develop a smart EMAS (SEMAS) with the help of IoT and GIS. The algorithm involves three procedures, where, the first procedure is devoted to generate a set of potential solutions (Hamiltonian paths through the houses). The second procedure is used for the proper movement of the solutions in the search space with the help of some well-defined perturbation techniques. If a route is not improved in this procedure using a selected perturbation rule then K-opt is used once for the same. The third procedure of the algorithm regenerates the stagnant solutions to overcome any local optima. The second and third procedures are repeated iteratively for fixing the optimal schedule. The effectiveness of the approach is tested using some test problems from TSPLIB and its efficiency, consistency, and accuracy are well established. The efficiency of the approach is also compared with some recently published heuristics on TSPs using statistical tests and its superiority compared to others is established. Using this heuristic, a case study in an urban setting is done to demonstrate the effectiveness of SEMAS.

A simple and efficient heuristic is designed for the Traveling Salesman Problem (TSP). A potential solution of a TSP is a permutation of nodes associated with it, and a shuffling of the positions of two nodes in the permutation is known as a swap operation (SO). Using SO, four perturbation rules are designed for searching a neighbor path of any potential path of the salesman. Similarly, cyclic crossover operation is used to design four more perturbation rules. The search process begins with a set of randomly generated potential paths(initial population). In each iteration, one rule is selected from this set of eight rules based on their performance to determine the perturbed (child) population from the parent population. To eliminate redundant paths, the population for the next iteration is selected from the union of parent and child populations. K-opt operation(for K=3) is also used to enhance any stagnant path. The algorithm is tested experimentally using several test problems from TSPLIB. The efficiency and consistency of the algorithm for solving large-sized TSPs are verified. Statistical studies are performed to check the performance of the approach concerning state-of-the-art heuristics, and the superiority of the proposed approach is established. Coordinating internet of things(IoT), GIS(geographic information system), Bluetooth technology, and scheduling strategies of the TSPs, several real-life problems can be dealt efficiently, e.g., emergency medical service, home delivery of online business, disaster management, etc. Using this approach, a model on urban health care for senior citizens is proposed and illustrated. As real-time estimations of different parameters are usually imprecise, proper methodologies are outlined to solve the TSPs involving imprecise cost matrices. The same approaches are used to demonstrate the model with fuzzy and rough data.