Vasavi College of Engineering

Self-localization is the capability of a wireless sensor network (WSN) to estimate the location coordinates of a given target node (TN), using the location knowledge of few anchor node (AN)s. The ANs are the sensor nodes installed with GPS modules, and their locations are known in prior. Precision improvement of the TN’s location is an important issue for effective data transmission in WSNs. Localization draws attention in location aware applications like nuclear attacks, object tracking, healthcare, supply chain management, biological attacks, and traffic monitoring, etc. In this article, a weight based AN selection strategy in localization algorithm for WSNs is proposed, which uses only four ANs where each one is installed with GPS module, and deployed within the sensing area. The weights are assigned to ANs depend on the distance between ANs and TNs to mitigate the effect of noise due to spatial and temporal changes in the sensing area. The proposed method uses improved cuckoo search algorithm, which is a nature inspired optimization technique to address the anisotropic nature of WSNs, and compute global optimal location coordinates of TNs. It also enhances the localization accuracy compared to the nascent localization algorithms in WSNs. Rigorous simulations are conducted to prove the efficiency of the proposed method using the performance metrics as number of ANs, mean localization accuracy, and communication range. Based on the simulation results, the localization accuracy of the proposed CERBLA algorithm is increased to 99.24% and the range measurement error is minimized to 1.18 m. The localization accuracy using the CERBLA algorithm is enhanced by 25.32%, 25.19%, 21.47%, 128.21%, 84.48%, and 181.35% compared to DCK-GWO, WOA-QT, DECPSODV-Hop, ECS-NL, QABA, and IDE-NSL-AWSN algorithms respectively. The performance of the CERBLA is outperforming the existing algorithms particularly, when the number of ANs are less than ten, and this feature is attractive to build cost-effective localization algorithms for indoor applications of WSNs.

This work presents a comprehensive investigation of performance and reliability enhancement in GaN/BGaN high electron mobility transistors (HEMTs) through dual trap engineering using copper (Cu) and iron (Fe) layers on a SiC substrate. The proposed device incorporates a GaN cap, n-type BGaN barrier, BGaN spacer, and GaN channel to facilitate strong polarization-induced 2DEG formation at the heterointerface. The introduction of Cu and Fe trap layers within the buffer region is systematically analyzed using TCAD simulations to evaluate their impact on electrical, RF, and reliability characteristics. The results demonstrate a significant improvement in device performance with trap engineering. The drain current (IDS) increases from 700 mA/mm in the conventional structure to 756 mA/mm and 784 mA/mm for Fe and Cu trap configurations, respectively. Similarly, the transconductance (gm) improves from 130 mS/mm to 140 mS/mm and 145 mS/mm. RF performance analysis reveals variations in gate-source and gate-drain capacitances, indicating enhanced charge control and reduced parasitic effects. Notably, the minimum noise figure is reduced from 10.7 dB to 10.2 dB in the Cu trap-based device, highlighting its suitability for low-noise applications. From a reliability perspective, the peak electric field is significantly reduced in the Fe trap structure (77,514 V/cm) and remains controlled in the Cu trap case (88,480 V/cm), compared to 122,304 V/cm in the conventional device. This reduction mitigates hot carrier effects and enhances device stability. The improved performance is attributed to enhanced carrier confinement, optimized 2DEG density, and effective suppression of leakage through trap-assisted field redistribution. Overall, the proposed Cu–Fe trap-engineered GaN/BGaN HEMT demonstrates superior electrical, RF, and reliability characteristics, making it a promising candidate for next-generation high-frequency and high-power electronic applications.

The widespread use of online learning has been facilitated by advancements in information technology. Less pertinent assessments and applications are available for students, nevertheless. All learning innovation efforts aim to improve education while simultaneously increasing student engagement in the classroom. By opting instructional resources that are suitable for each student’s learning preferences, teachers may increase the participation of their students in the teaching–learning process. This chapter aims to create and assess how well a learning style prediction based on AI model performs in a learning portal. In order to support the idea of personalized learning, an AI model was developed for the online learning site to provide educational resources that suit students’ learning preferences. Consequently, this project’s goal is to create a user interactive application that helps student understand his way of learning and improve himself in the way he understands and learns things.

Mortgage-backed securities (MBSs) have long been a staple in financial markets, offering investors exposure to the residential mortgage market. However, the prepayment risk associated with these securities presents a significant challenge for investors and issuers alike. Traditional models for predicting prepayment risk often fall short in capturing the complex dynamics of the mortgage market, leading to suboptimal risk management strategies. In this research paper, we propose a novel machine learning approach to predict MBS prepayment risk with the aim of enhancing risk mitigation strategies. Leveraging advanced techniques in data analytics and predictive modeling, our methodology integrates a diverse set of features, including borrower characteristics, economic indicators, and market trends, to generate accurate forecasts of prepayment behavior. Through comprehensive experimentation and validation using historical MBS data, we demonstrate the effectiveness of our approach in accurately predicting prepayment risk across various market conditions. Our results indicate significant improvements in prediction accuracy compared to traditional models, thereby empowering investors and issuers with valuable insights for making informed decisions and mitigating risk exposures. Our study contributes to the growing body of literature on MBS prepayment risk management by showcasing the potential of machine learning techniques to provide more accurate and reliable predictions, thereby enabling stakeholders to navigate the complexities of the mortgage market with greater confidence and efficiency.

Cloud computing has become the basis of modern computing, with on-demand provision of shared resources in an economical, scalable, and elastic manner. One of the key problems here is efficient scheduling of interdependent tasks with optimal mapping to virtualized cloud resources. The present paper proposes a novel scheduling algorithm called the Two-Phase Evolutionary Approach for Cloud Workflow Optimization (tp-EACO), which includes a user-defined preference at a well-defined stage in its optimization process. It thereby improves the alignment of generated solutions to particular user specifications. The new approach integrates two strategic phases: Preference Distance Strategy (PDS) for directing search to regions of preference and Preference Region Ranking Strategy (PRRS) for ranking and choosing optimal solutions. A mechanism for iterative improvement is provided through elite learning. tp-EACO, subjected to extensive experimental testing, demonstrates significant improvements in execution efficiency, cost savings, and usability of solutions over many scheduling algorithms.