Authors: Krishna Prisad Bajgai, Dr. Bhoj Raj Ghimire

Abstract: Early identification of unrecoverable loans is a critical requirement for financial institutions to maintain portfolio quality, comply with regulatory provisioning standards, and minimize credit losses. In Nepal, microfinance institutions and banks are mandated to report loan performance using the Nepal Rastra Bank (NRB) N002 monitoring framework, which contains borrower demographics, loan characteristics, delinquency behavior, and provisioning information. Despite the availability of structured regulatory data, most institutions continue to rely on rule-based aging mechanisms that fail to capture complex nonlinear risk patterns. This study proposes a machine learning-based framework for predicting unrecoverable loans using NRB N002-compliant datasets. A supervised classification problem is formulated, where loans are labeled as unrecoverable based on regulatory delinquency thresholds (Days Past Due >180 or Provision ≥50%). Three models—Logistic Regression, Random Forest, and Extreme Gradient Boosting (XGBoost)—are implemented and evaluated using recall, precision, F1-score, and ROC-AUC metrics, with special emphasis on recall to minimize false negatives in high-risk loan identification. Experimental results demonstrate that XGBoost achieves superior performance with near-perfect recall for unrecoverable loans and an ROC-AUC exceeding 0.97, significantly outperforming traditional statistical approaches. Explainability is ensured using SHAP-based feature attribution. highlighting delinquency duration, overdue principal, outstanding exposure, and provisioning ratios as dominant predictors. The findings confirm that machine learning models can substantially enhance early warning credit risk systems within Nepalese financial institutions while maintaining regulatory transparency and operational interpretability.

DOI: https://doi.org/10.5281/zenodo.18440690

 

Published by: vikaspatanker

Authors: Sahil Shinde, Pushkar Rahane, Sudarshan Suryavanshi, Krishna Tayde, Prof. Bhagawat S. Mohite

Abstract: This research Security is a major concern in homes, offices, and restricted areas. Traditional lock systems using mechanical keys have limitations such as key loss, duplication, and lack of access control. To overcome these issu es, this project presents the design and implementation of an RFID Door Lock using Arduino. The proposed system uses Radio Frequency Identification (RFID) technology to allow only authorized users to access the door. An RFID reader reads the unique ID of the RFID card or tag and sends it to the Arduino microco ntroller. The Arduino compares the scanned ID with the pre-stored authorized IDs. If the ID matches, the system.

 

 

Published by: vikaspatanker

Authors: Dr. Ruchita Sujai Chowdhary

Abstract: Sacred plants have long played an integral role in shaping ecological consciousness, ritual performances, and cultural identity within Indian society. Among these, Tulsi (Ocimum sanctum) and Peepal (Ficus religiosa) hold a distinctive presence as sacred, medicinal, and symbolic botanical entities embedded deeply in everyday religious and cultural practices. This qualitative study examines the Use, Knowledge, and Significance (UKS) surrounding these plants through community narratives in both rural and urban settings in North India. Utilizing narrative inquiry and ethnographic approaches, the research documents oral histories, lived experiences, ritual participation, and ecological perceptions expressed by diverse community members. Findings reveal that Tulsi and Peepal function not only as religious icons but also as powerful conveyors of environmental values, emotional wellbeing, and intergenerational continuity. Despite rapid modernization and urban transformations disrupting traditional practices, the enduring relevance of these plants demonstrates their potential as culturally grounded tools for ecological communication. The study argues that sacred plant traditions embody a form of “sacred ecology,” offering insights into sustainable cultural-environmental relationships and highlighting the need for preserving traditional knowledge systems.

DOI: https://doi.org/10.5281/zenodo.18430477

Published by: vikaspatanker

Authors: Ms. Chavhan S.S, Ms. Humbe P.D, Ms. Bodake S.S, Ms. Devkar M.R, Ms. Jadhav N.

Abstract: Street lighting plays a vital role in public safety but consumes a significant amount of electrical energy. Conventional street lights operate continuously throughout the night, leading to unnecessary power wastage. This paper presents the design and implementation of a solar powered smart street lighting system with motion detection to improve energy efficiency. The proposed system uses solar energy as the primary power source and a motion sensor to control light intensity based on human or vehicle movement. During periods of no motion, the light remains in dim mode, and it switches to full brightness when motion is detected. The system is controlled using a microcontroller and operates automatically without manual intervention. Experimental results show that the proposed system significantly reduces energy consumption and maintenance costs, making it suitable for smart city applications.

Published by: vikaspatanker

Authors: Okai Tettey-Antie Samuel

Abstract: Modern Quantum Key Distribution (QKD) often fails in highly dynamic mobile environments due to rigid post-processing architectures. This paper introduces a pioneering self-adaptive privacy amplification (SAPA) framework that replaces traditional static compression with a closed-loop controller. By integrating twelve distinct quantum noise models—including Non-Markovian and Gaussian Bosonic channels—we demonstrate that real-time entropy estimation can reclaim up to 25% of secure key material previously lost in mobile-induced fluctuations. Our results establish a new paradigm for "living" security in future 6G ecosystems.

DOI: https://doi.org/10.5281/zenodo.18429991

 

Published by: vikaspatanker

Authors: Nishant Kadam, Swarup Chaudhari, Rushikesh Patil, Hrishikesh Kakade

Abstract: Automated vehicle monitoring is a cornerstone of modern security infrastructure, essential for maintaining safety and operational efficiency in high-traffic environments such as industrial complexes, gated communities, and public facilities. Traditional manual surveillance methods are frequently plagued by human error, significant labor costs, and operational bottlenecks that compromise the integrity of security protocols. This paper presents a robust framework for an automated Vehicle Entry Monitoring System (VEMS) utilizing the state-of-the-art You Only Look Once (YOLO) object detection architecture. The proposed system integrates real-time video stream processing with advanced deep learning models to achieve high-speed detection and classification of various vehicle types, including cars, trucks, and motorcycles. A critical component of the methodology involves the integration of Optical Character Recognition (OCR) and tracking algorithms, such as DeepSORT, to automatically extract alphanumeric license plate data and maintain unique vehicle identities across consecutive frames. This integration enables the creation of a comprehensive, searchable database that cross-references detected plates with authorized whitelists for proactive access control. Experimental results demonstrate that the system ensures near 100% operational uptime by automating the data trail for security auditing and regulatory compliance. The framework provides a scalable solution for intelligent transportation management, significantly reducing manpower dependency while enhancing the reliability of entry logs. By combining real-time detection overlays with a centralized monitoring dashboard, this research offers a sophisticated, data-driven approach to facility security, fostering safer and more efficient urban mobility environments.

Published by: vikaspatanker

Authors: Aher Pratiksha Mahendra, Auti Samiksha Ankush, Adak Dnyaneshwari Santosh, Adinath shankar satpute

Abstract: This paper presents the design and implementation of an Alcohol Detection with Engine Locking system for vehicles using the MQ-3 alcohol sensor, HC-SR04 ultrasonic sensor, and Arduino UNO as the Master Control Unit (MCU). The system continuously monitors alcohol concentration in the vehicle cabin and automatically locks the engine if the alcohol concentration exceeds the predefined threshold level. The proposed system also incorporates a SIM900A GSM module to send alert messages regarding the vehicle's whereabouts to designated authorities. Additionally, the ultrasonic sensor measures the distance between vehicles and activates warning indicators when the safe following distance is compromised. Experimental results demonstrate that the system provides an efficient and reliable solution to control accidents caused by drunk driving.

DOI: https://doi.org/10.5281/zenodo.18429030

 

Published by: vikaspatanker

Authors: Hachimenum N. Amadi, Sopakiriba Maxwell West, Richeal Chinaeche Ijeoma

Abstract: The incessant national grid collapse has become a global embarrassment; from 2015 to May 2024 the Transmission Company of Nigeria (TCN) has recorded 105 cases of grid collapse. Phasor Measurement Units (PMUs) are necessary for the extensive use and efficient running of international power networks, the present Supervisory Control and Data Acquisition (SCADA) system used in Nigeria does not provide a robust and dependable solution that improves the power grid’s real time monitoring and control capabilities. PMU will reduce the frequency of power grid breakdowns and also resolve fault location troubleshooting safely and timely. The optimal placement of Phasor Measurement Units (PMUs) is an important requirement in power systems research, particularly for the localization of transmission line faults. This research has proposed a Modified Sparsity Genetic Algorithm Optimizer (MS-GAO) for optimal placement of Phasor Measurement Units (PMUs) in Power Systems over the standard Genetic Algorithm (GA) approach used in various related studies. To further validate the performance, the time complexity studies were performed to determine the better technique considering enhanced PMU placement. The proposed approach has been applied to two IEEE power system networks – the IEEE 6-Bus and 14-bus power networks. The simulations were performed using the MATLAB software tool and results compared with the standard Genetic Algorithm (sGA) on the basis of the percentage Classification Efficiency (CE) and the number of trial iteration runs (iters) used per simulation. The results showed that the proposed MS-GAO gave comparable CEs when compared to sGA with 100% CE for 100iters. However, it was found that reducing the iterations to about 50iters resulted in a degradation of CE. Thus, a compromise should be made between the number of iterations required and the level of CE needed in the problem solution. In addition, computational run-time complexity results considering the 6-bus power network revealed that the MS-GAO will give better run-times when compared to the sGA with an average run-time reduction of about 0.5s. Thus, it is recommended that the MS-GAO be employed for a higher power bus networks as the computational demands will obviously be higher using a sGA.

DOI: https://doi.org/10.5281/zenodo.18427464

Published by: vikaspatanker

Authors: Mr. Rushikesh Kalhale, Mr. Venkatesh Bansode, Mr. Utkarsh Maske, Prof.Deepa Shivshimpi

Abstract: Mathematics is at the base of all Artificial Intelligence (AI) systems. Throughout the AI lifecycle, mathematics is the pillar for representing data at the start, learning, reasoning on behalf of the human user and adapting in the mid-section, and finally optimizing any algorithm or data driven model at the end. This paper will discuss how the main mathematics will start to emerge as critical constructs for AI – linear algebra, calculus, probability and statistics, and optimization. We will demonstrate the pertinence of mathematical models as a pathway for the development of neural networks, machine learning algorithms, and data driven decision systems. In demonstrating examples of how mathematics has evolved as part of the responsive development of Artificial Intelligence, we can clearly delineate the ongoing, sometimes inescapable, role mathematics will have in defining intelligent systems in the future.

DOI: https://doi.org/10.5281/zenodo.18427209

Published by: vikaspatanker

Authors: Hachimenum Nyebuchi Amadi, Ogadinma Agha Onya,, Richeal Chinaeche Ijeoma

Abstract: The reliability of 33kV distribution networks is crucial to the stability of Nigeria's electricity supply. Serving as the interface between the transmission grid and 11kV feeders, these networks directly affect service delivery, customer satisfaction, and operational efficiency. This paper examines the operational challenges and reliability issues of the 33kV feeders within the Port Harcourt Electricity Distribution Company (PHEDC) network, with a focus on performance assessment using standard indices such as SAIDI, SAIFI, and CAIDI. Preventive maintenance, feeder automation, and improved operational practices are identified as key measures for enhancing reliability. Results reveal major network challenges such as overloaded feeders, poor voltage profiles, high technical losses, and frequent interruptions. Reliability indices, including SAIFI, SAIDI, CAIDI, and ENS, were significantly above IEEE and NERC thresholds, indicating poor service continuity. To address these deficiencies, the study proposes targeted improvement strategies such as feeder reconfiguration, installation of automated reclosers and sectionalizers, preventive maintenance, and upgrading of aging conductors and transformers. The study concludes that targeted investment in maintenance, automation, and workforce training can significantly reduce outages and improve service continuity.

DOI: https://doi.org/10.5281/zenodo.18427035

 

Published by: vikaspatanker