Priyadarsini K

Cluster head (CH) selection is a decisive control mechanism for prolonging the operational lifetime of Internet-of-Things (IoT)-enabled wireless sensor networks (WSNs) that underpin smart-city sensing. Although reinforcement learning (RL) has produced increasingly capable CH selectors, three weaknesses persist across the recent literature: the learned policies are opaque and therefore unsuitable for mission-critical operations; they are validated almost exclusively under stationary traffic that does not reflect urban traffic loads; and they ignore the energy-harvesting heterogeneity that characterises real deployments. This study presents eXplainable Energy-Aware Clustering Hierarchy (XEACH), an explainable, energy-aware RL framework that addresses all three limitations within a single design. We formalise CH selection as a constrained network-lifetime maximisation problem, establish the NP-hardness of its per-round subproblem, and recast the sequential decision process as a Markov decision process in which each node learns a state-conditioned value for cluster-head service and selection is driven directly by the learned policy. We prove the convergence of the learner under standard stochastic-approximation conditions and bound the per-round cluster-head count. To render the policy auditable, we attach a post hoc explainability module based on SHapley Additive exPlanations (SHAP) that attributes every CH decision to its driving features. The framework is evaluated on smart-city scenarios of 50–300 nodes, incorporating a non-stationary traffic model with diurnal congestion and stochastic event spikes, together with 30% energy-harvesting nodes. All results are averaged over multiple independent seeds and reported with standard deviations. Against LEACH, SEP, and HEED, XEACH extends first-node-death by 88% over LEACH at 100 nodes and consistently outperforms all three baselines across network sizes, with the margin over LEACH widening as the network grows. An ablation confirms that the reinforcement-learning policy, rather than any fixed heuristic, drives these gains: replacing the learned policy with a random policy collapses lifetime by an order of magnitude. The SHAP analysis of fidelity 0.87 identifies residual energy and distance to the base station as the dominant, physically coherent drivers of selection. We additionally calibrate the energy model to the measured datasheet parameters of a current LoRa transceiver to establish the real-time feasibility of deployment.

The security of today’s communication networks rests on public-key primitives—RSA and elliptic-curve Diffie–Hellman—that a large quantum computer would break with Shor’s algorithm, while the “harvest-now, decrypt-later” threat means traffic captured today is already at risk. Two defences exist but each is incomplete: quantum key distribution (QKD) offers information-theoretic security yet fails open under eavesdropping or channel loss, while post-quantum cryptography (PQC) is computationally secure and always available but lacks physical-layer guarantees. This paper presents a dual-layer, trust-aware key-management framework that uses BB84 QKD as the primary channel and a Ring-LWE/Kyber key-encapsulation mechanism as a quantum-safe fallback, governed by a controller monitoring the quantum bit-error rate (QBER) and a trust signal. When intercept-resend eavesdropping pushes the QBER past the 11% abort threshold—reached at roughly 40% interception—the controller falls back to PQC, holding secure availability at 100% where a QKD-only system degrades to 78.5%. The lattice fallback is cheap ( ≈ 0.47 m per operation) and Kyber-768 gives 192-bit quantum-safe security. Coupling the fallback and re-keying rate to a node trust score integrates physical-layer and computational security into one always-available, forward-secure key-management layer. To our knowledge, this is the first key-management framework in which a single network-layer trust signal jointly governs QKD/PQC layer selection and the re-keying cadence, unifying physical-layer eavesdropping detection, computational quantum-safety, and forward secrecy in one cross-layer policy.

—Pregnancy-related complications remain a lead- ing cause of maternal and neonatal mortality, particularly in resource-constrained settings where timely clinical intervention is often unavailable. Identifying women at elevated risk before complications escalate is therefore a pressing need. This paper presents a multi-class classification framework that predicts maternal risk levels — Low, Mid, or High — from routine physiological measurements: age, blood pressure, blood glucose, body temperature, and heart rate. Working with a dataset of 1014 records, we applied label encoding, StandardScaler normalization, and stratified splitting as part of preprocessing. Three classifiers were trained and compared: Random Forest, Support Vector Machine, and XGBoost. After hyperparameter tuning, the XGBoost model reached 86.7% accuracy, a cross- validated Macro F1-score of 0.82, and a high-risk class recall of 0.88. To make predictions interpretable for clinical users, SHAP values were incorporated to expose the contribution of each feature to individual risk decisions. Index Terms—Maternal Health, Machine Learning, XGBoost, Risk Prediction, Explainable AI

This approach of speech emotion recognition(SER) is a problem in the subject of deep learning. The primary objective of this study is to identify a variety of feelings based on the input of audio utterances as well as the Rnn networking. We used the long short-term memory networks (LSTM) model because it deals with speech features and aims to improve speech recognition rates. Initially, this method of speech emotion was done by machine learning approach but recently deep learning techniques are proven to be an alternative to SER providing better accuracy.We used the dataset's audio files as input, performed feature extraction, and trained the model. Considering researching the published works from prior periods' research and putting the suggested models through their paces on our own datasets, We expect that the overall file classification efficiency will be 99.73

During these modern days everything revolves around technology, thus everything is getting digitized. From teenagers to working age groups use technology, so it is important for computers to classify digits based on handwriting. Many of them have different handwriting which sometimes cannot be identified by humans themselves, so what we have to do is by using machine learning we could make the computer read with an accuracy of 95%. This will surely benefit the process of storing old documents without retyping everything as it will consume a lot of time, instead they can take a picture of the documents and upload into the system which will in turn go through it and show it in text form. On this topic we have done our share of research, and also we have implemented it practically. All types of handwritten documents can be read with a high percentage of accuracy, which makes many people's lives easier. The project is still in its development and early stages, but once it reaches its destination it could be very useful.

The wireless sensor network (WSN) consists of a large number of randomly distributed nodes capable of detecting environmental data, converting it into a suitable format, and transmitting it to the base station. The most essential issue in WSNs is energy consumption, which is mostly dependent on the energy-efficient clustering and data transfer phases. We compared a variety of algorithms for clustering that balance the number of clusters. The cluster head selection protocol is arbitrary and incorporates energy-conscious considerations. In this survey, we compared different types of energy-efficient clustering-based protocols to determine which one is effective for lowering energy consumption, latency and extending the lifetime of wireless sensor networks (WSN) under various scenarios.