Koushik Roy

Despite abundant sunlight, vitamin D deficiency remains a significant health challenge in countries like Bangladesh, driven by dietary habits, socioeconomic factors and limited sun exposure. This study leverages machine learning (ML) to predict vitamin D levels in infants aged 12–24 months, using real‐world data from the International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B). Data preprocessing included outlier removal via the ‐score method and Min–Max normalization for variable standardization. Ten ML models, including random forest, XGBoost, gradient boosting and Gaussian regressor, were evaluated with hyperparameter tuning to optimize performance. A stacking ensemble model, combining Gaussian regressor and extra trees as base learners and XGBoost as the meta‐learner, delivered the best results (MSE: 10.971, MAE: 2.479, RMSE: 3.312, : 0.933, MSLE: 0.005, adjusted : 0.918). Evaluation metrics such as MSE, MAE, MSLE, RMSE, , and adjusted confirmed the stacking ensemble model's superior predictive accuracy. SHAP‐based interpretability analysis on the stacking ensemble identified total (total time), total (sun exposure) and UV index as the primary predictors, enhancing clinical transparency and trustworthiness of the model. This research highlights the potential of ensemble learning in healthcare and provides a foundation for early intervention strategies to address vitamin D deficiency.

The simultaneous achievement of scalability, security and decentralisation remains an open problem for distributed ledger technologies. This paper introduces InternxtChain, a novel framework leveraging Internxt's decentralised storage infrastructure with zero‐knowledge proofs (ZKPs) and sharded proof‐of‐storage (SPoS) consensus. Specifically, erasure‐coded sharding ensures data availability and fault tolerance by splitting files into encoded fragments distributed across nodes; BLS‐381 aggregated signatures enable efficient consensus by compressing multiple signatures into a single short proof; and zk‐SNARK audits provide tamper‐evident storage verification without revealing user data. InternxtChain addresses this challenge through three synergistic mechanisms: (i) erasure‐coded sharding with (6,3) Reed–Solomon encoding, (ii) zk‐SNARKs for storage auditability and (iii) an SPoS consensus based on BLS‐381 aggregated signatures. Experimental evaluation on a testbed of 2048 nodes across 16 geographic regions shows that InternxtChain processes 2800 transactions per second (TPS) with a median latency of 420 ms, while maintaining 99.9% data integrity under up to 30% Byzantine nodes. These results establish a practical path toward harmonising Web3 principles with real‐world throughput, cost and General Data Protection Regulation (GDPR) auditability requirements.

Contemporary distributed ledgers face an inherent trilemma when attempting to simultaneously optimize three critical properties: transaction throughput (measured in transactions per second [TPS]), network security, and node decentralization. Current decentralized storage platforms such as Filecoin encounter constraints in managing peak transaction loads effectively. We propose a novel storage architecture that integrates Arweave’s permanent data layer with an enhanced proof‐of‐access (PoA) consensus protocol and adaptive block management. While control‐theoretic methods are established, their novel integration with Arweave’s permanent storage and PoA consensus creates a unique symbiotic relationship that directly tackles the blockchain trilemma. Through transaction metadata compression to 48‐byte entries, our framework achieves a transaction density of 7200 entries per megabyte of block capacity. Under baseline conditions (1 MB blocks generated every 44 s), this yields 162 TPS, scaling to a simulated peak of 7200 TPS with 200 MB blocks under optimal network parameters. Experimental validation through simulation and a multiphase global testbed (25–112 nodes) demonstrates robust security characteristics with 0.82 chain quality under 40% adversarial presence, defense mechanisms against prevalent network attacks, and 99% reduction in annual storage growth compared to conventional blockchain systems, outperforming modern alternatives including Sui (5.3× storage efficiency) and Avalanche (3.5× storage efficiency). Game‐theoretic proofs establish formal security guarantees, while extended adversarial modeling confirms resilience under worst‐case conditions. To ensure reproducibility and foster further research, all simulation code and deployment tools for the testbed have been made publicly available on GitHub. These findings present a viable equilibrium between essential blockchain properties for practical implementations.

The persistent blockchain trilemma—balancing scalability, security, and decentralization—hinders distributed systems adoption. This paper introduces StorjLedger , a novel framework integrating Storj's decentralized storage with Proof‐of‐Storage consensus to resolve this challenge. By partitioning data into 256‐byte erasure‐coded shards distributed across 84 global nodes, StorjLedger achieves 98% storage redundancy while compressing on‐chain metadata to 64 bytes per transaction. The system attains 980 TPS throughput with 1.2‐second latency and maintains 97.9% retrieval success under Byzantine faults—outperforming Filecoin by 32% while reducing storage costs by 76% versus conventional blockchains. Live testing across seven regions confirms robustness, with ≤ 3.2% retrieval failures and 95.4% quorum consistency during adversarial churn.