The demand for data-driven insights in government has highlighted the importance of collective analytics. This study attempts to explore the key challenges of collective analytics in the context of Indian e-governance and the framework for addressing them. The study is based on a literature review, references to two cases, and expert views obtained from professionals involved with analytics solutions in government. In this study, analytics projects are considered as dashboard-based analytics. Based on the content analysis of expert responses, 14 key challenges of collective analytics in e-governance have been identified. The novelty of the present study is the focused exploration of challenges and their framework related to collective analytics in e-governance-a topic that received limited attention in the extant literature. This study brings forth the fact that unless the challenges of collective analytics in e-governance, including those related to data visualization, data quality, capacity building, technological capabilities, and inter-agency communications, are recognized, the implementation of collective analytics can be challenging. This study provides the basic understanding needed for data-driven governance through collective analytics. The output of the study will be helpful to the managers, e-governance experts, academicians, planners, and policymakers to understand the dynamics of collective analytics in government for handling discussed challenges well in advance. This study will also helpful to reduce the cost and time of the collective analytics project for effective decision-making.

The research objective is to formalize the cognitive stratified framework of digital fiscal control through the unification of information and analytical tools based on decomposition, topological analysis, and Unified Modelling Language (UML). The study employed the following methods: SWOT analysis of solutions for the digitalization of fiscal tax control systems, decomposition and range analysis of digitalization technologies, topological analysis of digital information and analytical tools, and UML modelling of the framework for the digitalization of fiscal tax control systems. The developed framework is presented as a conceptual and architectural design structure that integrates artificial intelligence (AI)/machine learning (ML) risk stratification, Distributed Ledger Technology (DLT) traceability, autonomous compliance, and P2P interoperability to outline architectural integrity and procedural resilience as intended design properties rather than empirically demonstrated effects. SWOT, decomposition and range analysis, as well as topological analysis supported the identification of a unitary routing logic for risk, compliance, and verification flows that may contribute to fiscal transparency and evasion risk mitigation under subsequent pilot testing. The academic novelty is associated with the systemic identification, decomposition, structured organization, and topological mapping of information-analytical tools of fiscal control, which enabled the formal representation of a cognitively stratified architectural and functional topology of a digital fiscal control framework using UML modelling.

Accurate option pricing is critical for the effective functioning of financial markets, providing traders and investors with the means to hedge risks and capitalize on market movements. Traditional models such as the Black-Scholes, Binomial Tree, Trinomial Tree, Monte Carlo Simulation, and the Garman-Kohlhagen model have long been the standard for option pricing. However, these models often face limitations in capturing market complexities and extreme events. We propose here a hybrid approach that combines Genetic Algorithm (GA) optimization with Backpropagation (BP) neural networks to enhance the precision of option pricing. It uses HS300 index stock data from 2013 to 2022, including stock prices, volumes, and price changes. The hybrid GA-BP model is tested for its ability to make more accurate price predictions. The model helps investors make better decisions by improving pricing strategies and managing risks effectively. The Hybrid GA-BP neural network model leverages the global search capabilities of GA to optimize the initial weights and biases of the BP neural network, thereby avoiding local minima and improving convergence rates. This integrated model is trained and tested on historical market data, with its performance benchmarked against traditional models. Empirical results demonstrate that the Hybrid GA-BP neural network model significantly outperforms traditional models in terms of pricing accuracy. The model shows superior precision when comparing actual market prices with predicted prices, reducing errors and increasing reliability. This enhancement in pricing precision can lead to more informed trading decisions and better risk management strategies. The findings of this research contribute to the growing body of knowledge in financial engineering by showcasing the potential of hybrid machine learning approaches in financial modeling. The Hybrid GA-BP neural network model presents a promising tool for practitioners and researchers aiming to improve option pricing methodologies in increasingly complex financial markets. 

This research examines RAFA-BioAuth, a risk-adaptive, fairness-aware framework for mobile banking in cases of presentations and facial occlusions. The proposed solution combines aspects of: Identity Similarity, Passive Presentation Attack Detection (PAD), Asymmetric Financial-Risk Estimation and Fairness Regularization at the Identity Level. For evaluation purposes, all benchmark datasets were split into subject disjoint training, validation, and testing sets. The threshold values from the validation set were used. Bootstrap resampling was employed to estimate the variance. Monte Carlo simulations were performed to estimate the risk. The results showed that identity verification (AUC = 0.548) and PAD (AUC ≈ 0.55) were poor. Comparing RAFA to the AND rule resulted in FAR = 0.20, FRR = 0.31, and expected risk = 340. On the other hand, the AND rule had lower FAR of 0.09, but increased FRR to 0.78. Finally, a conservative end-to-end approach produced an FRR of 0.686. Thus, our results indicate trade-offs rather than production-readiness as we did not perform deployment, cross-device, or cross-dataset validations on our solutions. We present the contributions of this research as being an interpretable integration and not a new algorithm.

Technology Business Incubation (TBIs) has become a global phenomenon integral to the growth of regional innovation and startup ecosystems. The availability of high-quality infrastructure and facilities lays the foundations of the entire startup ecosystem for providing essential support services that directly impact entrepreneurial success. The incubation capacity of TBIs across different regions can foster competition and collaboration among these regions, provide avenues for enhancing enterprises’ incubation capabilities, and assist entrepreneurs in assessing the strength of regional incubation. However, with their rapid expansion, the performance evaluation also becomes increasingly complex due to the diversity of converging factors such as complex technologies, varying nature of relationships of VCs, and entrepreneurial competencies of the founders incubating startups at the TBIs. Traditional Machine Learning performance evaluation and prediction models struggle to capture these dynamic variables, while also suffering from privacy vulnerabilities, low accuracy, and reliance on centralized third parties. This often leads to single points of failure, performance bottlenecks, and sometimes increased costs. To address these challenges, we employed Privacy-Preserving Federated Learning with Blockchain (PPFL-BC), a novel framework designed for improving the mechanism of performance measurement and prediction for remote TBIs while ensuring that the privacy of entities and the data remains secure. We utilize capabilities of Artificial Neural Network (ANN) and gradient boosting-enabled federated learning to train the model of each TBI locally. In the process, no private and sensitive business data is shared outside the network, significantly reducing the risk of privacy breaches. Besides this, all the locally trained models are aggregated into a unified predictive model at the central aggregation unit, which ultimately improves the overall accuracy of the performance prediction mechanism for the entire population of TBIs. In our model, the decentralized blockchain network is also used to address security concerns related to unauthorized access and data manipulation thereby ensuring transparent and tamper-proof model updates. We evaluate the performance of our proposed PPFL-BC model by utilizing real-world business incubation datasets. The simulation results show that our model outperforms the centralized performance prediction models in terms of accuracy, precision, recall, and F1-score. The results show that the proposed PPFL-BC model outperforms benchmark models with an accuracy of 84% and precision of 0.92, which shows the efficiency and reliability of our model in predicting and validating TBI success rates.

This paper proposes IPAMS (Interview Performance Assessment using Gen AI), which is an AI-driven platform that automates interview evaluations using advanced technologies like Convolutional Neural Networks (CNN) to gain insights on facial emotions and expressions, Large Language Models (LLM) to generate and process interview questions, YOLO (You Only Look Once) for real-time object detection, and APIs for speech-to-text transcription and behavioral analysis. The system captures video responses and analyzes key elements such as sentiment, speech patterns, body posture, and facial expressions, generating a detailed report. This report highlights a candidate’s strengths and areas of improvement and is sent directly to their email with actionable insights. IPAMS modernizes recruitment by providing unbiased assessments, saving time and resources for recruiters. For candidates, it offers a valuable mock interview tool, delivering feedback on technical skills, confidence, stress levels, and nonverbal communication. By combining cutting-edge AI and analytics, IPAMS delivers an efficient, objective, and insightful solution for recruitment and self-assessment, benefiting all stakeholders in the interview process.

Job Shop Scheduling Problem (JSSP) has become one of the key issues in a contemporary manufacturing system in which the task is to optimally schedule jobs to the machines to reduce the time and resources used in production. Good scheduling is critical in enhancing the productivity and competitiveness of manufacturing industries. In this research, Artificial Fish Swarm Optimization (AFSO) algorithm is used to optimize the JSSP in minimizing makespan, total work load and maximum work load in machines. The AFSO strategy models the swarm behaviour of fishes to search and forage the search space in an efficient manner to prevent its early convergence to local optima. The model incorporates a disturbed state in the world to improve the direction in search and the speed of convergence. The effectiveness of the suggested AFSO method is compared and tested with the traditional and sophisticated optimization algorithms like Particle Swarm Optimization (PSO) and Artificial Bee Colony (ABC) methods. The experimental findings prove that the offered technique provides better results in convergence rate and solution quality. The results prove that AFSO is a useful and promising method of solving complex problems in production system scheduling.

The article discusses methods for averting irrational product placement in warehouses. This problem is relevant for many manufacturing enterprises and trade organizations. The most common embodiments of the problem are unoccupied areas, or a lack of free storage space, difficulty in locating a specific product, and challenges in shipping it from the warehouse. All this leads to unnecessary costs for the business entity and hurts its financial and economic activities. The author suggests an integrated approach to warehousing. It integrates both classical optimization models and iterative approval procedures for accounting for the human factor. The key criteria in this case are minimizing costs, the cargo flow in the warehouse, and maximizing the utilization factor of the usable area. The optimal placement of goods is to achieve maximum compression of their residues in the warehouse while minimizing their movement. The presence of two contradictory criteria makes the task a task of consistent optimization. The article discusses the possibilities for solving the optimization problem when conflicting target criteria and differing preferences are present. We are using the example of a storage room with 16 racks for water heaters and similar equipment. As a result of matching optimization procedures, it was possible to reduce the average cost of moving goods by 7.1% and increase the free warehouse area by 15 times. We performed the experiments over the seven days of the warehouse’s operation. The practical value of the research is that, through this approach, we find a compromise in conditions of conflicting opinions and interests.

Remote sensing images are complex, which makes it difficult to interpret and generate semantically appropriate textual description. To get a semantically relevant description, it is important to identify complex objects and understand the contextual relationships between them. In such cases, deriving contextually accurate information while maintaining semantic coherence is challenging. Therefore, a specifically designed model architecture is required to generate semantically relevant descriptions. This paper discusses a deep learning-based approach to generate remote sensing image descriptions using an end-to-end encoder-decoder model with soft attention. The UC Merced (UCM) dataset is used for training, which includes multiple captions per image capturing various scene aspects. To further assess the robustness and generalizability of the proposed approach, its performance is additionally evaluated on more complex datasets such as RSCID and Sydney Captions. This study presents an end-to-end CNN–LSTM encoder–decoder framework enhanced with soft attention for semantic description generation from remote sensing imagery. The framework employs a VGG16 encoder to extract a 4096-dimensional visual feature vector, which is projected into a 256-dimensional representation and processed by a 256-unit LSTM decoder. The soft attention mechanism dynamically computes attention weights using the encoder features and decoder hidden state, enabling the model to emphasize relevant visual information during word generation. Multiple CNN encoders and learning rates are evaluated with LSTM decoders, both with and without attention, on the UCM, RSCID, and Sydney Caption datasets. At a learning rate of 0.0001, VGG16–LSTM with soft attention achieves BLEU-4 (B4) scores of 0.6636, 0.6636, and 0.5864 on the UCM, RSCID, and Sydney Caption datasets, respectively, compared with 0.1643, 0.1647, and 0.1745 for VGG16–LSTM without attention. The results demonstrate that soft attention substantially improves description generation by strengthening visual–linguistic alignment and enabling more contextually relevant and semantically coherent descriptions across datasets with varying scene complexity.

Advanced Metering Infrastructure (AMI) connects smart meters, data concentrator units, and utility control centers through persistent two-way communication. This architecture improves demand response and distributed-energy management, but it also exposes resource-constrained meters to replay, false-data injection, physical extraction, and long-term key compromise. This article develops a formally verified and statistically evaluated lightweight AMI authentication and key agreement protocol for resource-constrained smart-grid deployments. We first reconstruct the AMI authentication workflow as a four-message lightweight authenticated key exchange and map each entity, message, and key dependency to a smart-grid deployment model guided by NISTIR 7628 and IEC 62351. We then identify replay-within-window exposure, insufficient responder freshness, weak identity-to-key binding, missing key-compromise impersonation protection, and retrospective session-key recovery. To address these weaknesses, we propose AMI-AKE, a transcript-bound protocol using ephemeral Curve25519 contributions, session identifiers, nonce and timestamp binding, binding signatures, and separate key-derivation function (KDF) outputs for encryption and integrity. ProVerif-style verification queries and an extended Canetti-Krawczyk (eCK)-oriented game proof are provided for mutual authentication, secrecy, forward secrecy, and key-compromise impersonation (KCI) resistance. A Contiki-OS and ARM Cortex-M4 benchmark with 1,000 repeated trials reports 18.4 +/- 1.2 ms authentication latency, 542 +/- 9.1 sessions/s throughput, and 99.2 +/- 0.4% false-data-injection detection under controlled prototype conditions. The proposed design replaces subjective security labels with objective metrics, confidence intervals, and a reproducible simulation plan for 1,000-10,000 smart meters.

The paper deals with coalitions whose members are unselfish. Coalition members do their best to complete the arising tasks, and do not expect to receive a reward. A coalition member can be an entity such as a social or governmental organization, a military unit, or a complex technical device such as an autonomous robot, or any other entity that has the capabilities, willingness, and ability to cooperate. The paper considers the non-redundant coalitions, which have only those coalition members without whom they cannot perform the tasks. In the paper, we consider only the situations when substitution of failed coalition members is impossible. A coalition tolerates the failure of its members by using the surplus of coalition capabilities. In our research, coalition capabilities are understood as resources and services (e.g., materials, energy, power etc.). During the execution of tasks, one or more members of the coalition may fail. The paper uses the probability of the event that the coalition tolerates the failures of its members to evaluate the coalition fault tolerance. A method is proposed for determining the coalition fault tolerance in the case of multiple member failures. Complexity of the proposed method and its applicability are assessed. 

A system that is personalized and capable of automatically suggesting appropriate courses based on a user's particular questions and interests provides customized recommendations. The system employs an LTR model relying on XGBoost to learn relationships between the queries and the courses. Dynamic ranking feature refinement enhances ranking, and a feedback loop constructs incrementally improving the recommendations by applying relevant courses to update the model. The scraped educational sites are the foundation of the dataset, where there are granular course descriptions as well as the interaction logs. Evaluation results indicate that the system is able to produce high ranking outcomes as evidenced by an NDCG value of 0.85 and high values for MRR. The system is able to produce low query processing latency, making it possible for real-time responsiveness. User feedback analysis following system retraining indicated a 90% increase in user satisfaction. The suggested framework is dynamic and provides personalized recommendations for courses in various learning environments.