G. Pullaiah College of Engineering and Technology

Knee osteoarthritis (OA) is a popular and progressive degenerate joint disease and require a predictor of risk at the primary stage to predict drug treatment and propose an early intervention. The current methods of deep learning are not sufficient enough to deal with integrating the heterogeneous modality component, modeling of irregular temporal follow ups, and capture of the fine grained variations of pathological types. In order to address such limitations, this paper suggests a new multimodal approach named, RAD-OAEnsemble which introduces three main innovations. First, a novel Cross Bi-modal Heterogeneous Graph Fusion Network (CBHGF-Net) is introduced for the multi-scale semantic fusion of Radiographs and Electronic Health Records (EHRs). Second, a Modified DenseNet with Multi-scale Attention Criss-Cross Blocks (MACC-Blocks) is presented to extract the anatomical features of knee OA patterns by utilizing dilated convolution, criss-cross attention, residual attention gating. Third, a Dual-Stream Tranfo-Convo (DTC) Deep Temporal Embedding module is proposed that consists of a Time-Warped Convolutional Blocks and a Temporal Decay Encoded Informer for the knee OA risk prediction. The suggested approach delivered better outcomes with accuracy of 99.275%, precision of 98.175%, specificity of 99.165%, sensitivity of 98.95%, and F1-score of 98.175% in having considerably a low rate of false positives (1.875%) and false negatives (2.175%). The findings support the relevance of the suggested framework in providing resilient, time-sensitive, and explainable OA progression predictions by multimodal combination and high-fidelity feature learning.

High population mobility and imbalances of housing markets are the critical challenges to sustainable urban and regional planning. The traditional forecasting models tend to focus on demographic change and housing pressure as independent aspects, hindering their capacity to capture interdependence of the two aspects. The current paper suggests a multi-task spatio-Temporal learning model that closely predicts the migration pressure and housing stress with the help of publicly available macroeconomic and demographic statistics on a national scale. Composite indices that capture the pressure of migration, the tightness of houses, and the affordability risk are obtained and as such, they are sourced and fed into a common Long Short-Term Memory (LSTM) architecture with two prediction heads. Experimental analysis using a country-level panel data has good predictive power, especially among housing stress predictions, where the model has a mean absolute error of 0.718, root mean squared error of 0.888, and a positive coefficient of determination (R2 = 0.269), which is better than statistical and single-task baselines. The prediction of migration pressure is still competitive with an MAE of 0.779 and an RMSE of 1.080. Further experiments involving cross country transfer reveal that the representations learnt in common can be used with predictive power in varying national states. These results demonstrate the usefulness of integrated modeling in the context of migration-housing interactions and give a facts-based basis on which to anticipate the appearance of new housing stress patterns at the national level at an early stage.

Cases of fire in residential, industrial, and forest areas are still very dangerous to life and infrastructure because there is delayed detection and false alarm systems. A real-time smart fire detection and prevention system, which combines IoTbased multi-sensor networks with machine learning algorithms to achieve high-precision early warnings, are presented in this paper. The suggested framework uses temperature sensors, smoke sensors and gas sensors that are implemented on an edge-enabled IoT platform to continuously obtain environmental data. A trained supervised machine learning model is used to differentiate normal fluctuations and critical fire conditions, thus minimizing false alarms greatly. The system also includes a real-time visualization and emergency response coordination system powered by cloud-based assistance. Through experimental analysis, it is shown that the proposed method has high detection accuracy, low latency, and sound fire classification in a variety of operating conditions. Smart prevention module activates suppression measures and emergency signals automatically, thus providing an opportunity to suppress fire hazards within a short period of time. The findings reinforce the fact that the proposed AI-enabled IoT system can be used as a scalable, economical, and robust solution to next-generation smart fire safety applications using smart buildings and industries.

Although large language models have been well-suited to a range of natural language processing tasks, adapting them for domain-specific cloud applications in low-resource settings is difficult because they require substantial memory, training data, and hardware resources. Fine-tuning these parameters at the right cost is important, but existing practices like LoRA, adapters, and prefix-tuning use fewer trainable parameters and still need to be optimized for deployment under dynamic cloud resource constraints. This paper proposes a resource-aware quantized parameter-efficient fine-tuning framework for adapting large language models in low-resource cloud environments. The proposed framework integrates quantized low-rank adaptation with a lightweight resource-monitoring mechanism that dynamically selects the appropriate fine-tuning strategy based on GPU memory, training time, latency, and estimated cloud cost. Experimental evaluation is conducted using representative NLP datasets and simulated cloud-resource settings. The results demonstrate that the proposed approach reduces trainable parameters, memory usage, and training cost while maintaining competitive accuracy compared with full fine-tuning and conventional PEFT methods.

The emergency departments are still experiencing challenges in terms of operational matters arising from overcrowding and the lack of information on the availability of hospital beds and blood resources. In emergency situations, any delay in the identification of suitable hospitals and resource allocation may have a negative impact on the health of patients. This paper introduces a real-time intelligent system for emergency response and hospital bed and blood resource management. The system consists of an emergency severity identification module based on deep learning models that process triage inputs and Based on the severity level, the system assists in allocating priority to hospital bed and blood resource allocation using a centralized dashboard. In addition, an AI-based chatbot and voice assistant is incorporated to facilitate faster engagement and basic emergency services. Simulation-based experimental analysis on emergency data sets shows that the proposed system outperforms existing machine learning-based triage systems in terms of classification accuracy. System-level analysis also shows faster response time and enhanced efficiency in hospital selection during emergencies.