As Crime rate is increasing in many parts of the world whereas law enforcement agencies have limited staffing resources. This is a serious problem and can be addressed using intelligent computation techniques. This research work is focused to improve crime prediction as well as allocation of Police (cops) as per their efficiency and nature of crime. This paper explores a hybrid approach that combines unsupervised machine learning and generative AI to address both of these issues. It is done by analyzing historical crime records, looking at when, where, and what kinds of incidents occur, along with whom they affect. Clusters are uncovered using the analytics to hint at deeper patterns beneath the surface. These groupings help to predict not just the likelihood of crimes in certain areas, but also the nature of those events. This problem definition is further improvised by integrating a Large Language Model (LLM) that can interpret natural language inputs like incident descriptions and match them with the backgrounds of available officers. This way, the system suggests more thoughtful deployments, aiming to send the right person for each job, whether that’s a case of domestic violence or a fast-moving theft. We apply this method to a real-world dataset from Los Angeles, which poses additional challenges like incomplete data on officer capabilities. Even so, by combining the insights from clustering with the interpretive power of generative AI, we generate informed, skill-sensitive deployment recommendations. This blend of data science and contextual AI offers a practical step toward smarter, more responsive policing, tailored not just to statistics, but to the people behind them.
In this study, we develop and analyze a tri-trophic predator prey model that incorporates predator-induced fear in the prey population and wind-driven modulation of predation efficiency. The model represents interactions among insects (prey), birds (predators), and feral cats (top predators), where fear induces non-consumptive effects that suppress prey growth, while wind acts as an environmental disturbance influencing intermediate hunting efficiency. We establish the positivity, boundedness, and existence of biologically feasible equilibria and derive sufficient conditions for both local and global stability using Jacobian analysis and Lyapunov functions. Conditions for saddle-node, transcritical, and Hopf bifurcations are obtained, revealing critical thresholds that govern stability loss and the emergence of oscillatory dynamics. Comprehensive numerical investigations including bifurcation diagrams and iso-spike patterns reveal a wide range of dynamical behaviors, from steady states and periodic oscillations to chaos. The presence of chaos is confirmed through the computation of maximum Lyapunov exponents. To suppress chaotic oscillations, a linear feedback control strategy is designed around the interior equilibrium, and its effectiveness is verified analytically via eigenvalue analysis and numerically through time-series simulations and control-parameter bifurcation diagrams. Global sensitivity analysis based on Partial Rank Correlation Coefficients (PRCC) identifies the most influential parameters governing long-term system dynamics. In addition, a Physics-Informed Neural Network (PINN) framework is employed for robust parameter estimation, representing a novel application of PINNs to chaotic ecological systems. The results indicate that increased predation fear has a stabilizing effect by suppressing chaos, whereas strong wind flow tends to destabilize the system by promoting oscillatory and chaotic dynamics. PINNs method found four unknown parameters with only 0.51% error on average. The neural network predictions matched the true population dynamics with R2 = 0.99, meaning it explained 99% accuracy. This worked 100 to 500 times faster than traditional optimization methods. Overall, the study highlights the combined roles of behavioral responses, environmental disturbances, control strategies, and machine learning tools in shaping complex ecological dynamics.
Emphysema, a primary component of chronic obstructive pulmonary disease (COPD), causes progressive dyspnea through the destruction of alveolar membranes. This structural degradation reduces the available surface area for gas exchange, significantly impairing oxygen delivery to the bloodstream. While oxygen therapy is a critical intervention, the inherent physiological complexities, specifically transit time delays and dynamic respiratory demands, make precise oxygen regulation exceptionally difficult. To address these challenges, this study develops a comprehensive mathematical model of the emphysema-affected respiratory system, incorporating specific parameters for time delays in oxygen exchange. A novel Intelligent Set-point Modulated Fuzzy Model Reference Adaptive Controller (SFMRAC) is proposed to enhance oxygen regulation. This control architecture advances traditional Model Reference Adaptive Controller (MRAC) by integrating a normalization factor, fuzzy logic tuning, and set-point modulation. This hybrid approach allows the system to adapt to nonlinear physiological variations and maintain stability despite the transit delays characteristic of damaged pulmonary tissue. The effectiveness of the SFMRAC was evaluated through a simulation study conducted in MATLAB/Simulink. Results demonstrate that the proposed controller provides superior tracking performance and robustness compared to MRAC, particularly when subjected to varying set-points and significant exchange delays. The results suggest that the SFMRAC offers a promising computational framework to improve the automated delivery of oxygen therapy in clinical settings for COPD patients.
Most existing medical dialogue systems operate in a single-turn question–answering paradigm or rely on template-based datasets, limiting conversational realism and multilingual applicability. We introduce IndicMedDialog, a parallel multi-turn medical dialogue dataset spanning English and nine Indic languages: Assamese, Bengali, Gujarati, Hindi, Marathi, Punjabi, Tamil, Telugu, and Urdu. The dataset extends MDDial with LLM-generated synthetic consultations, translated using TranslateGemma, verified by native speakers, and refined through a script-aware post-processing pipeline to correct phonetic, lexical, and character-spacing errors. Building on this dataset, we fine-tune IndicMedLM via parameter-efficient adaptation of a quantized small language model, incorporating optional patient pre-context to personalise multi-turn symptom elicitation. We evaluate against zero-shot multilingual baselines, conduct systematic error analysis across ten languages, and validate clinical plausibility through medical expert evaluation.
This study provides a comparative evaluation of three state-of-the-art large language models (LLMs), namely OpenAI’s (San Francisco, CA, USA) GPT-4.0, Google’s (Google LLC, Mountain View, CA, USA) Gemini 2.0 Flash, and Meta’s (Meta Platforms, Menlo Park, CA, USA) LLaMA-4-Scout-17B-16E, in a decision-oriented framework in which the models generate structured outputs based only on historical closing-price data. The evaluation covers 150 stocks sampled from three countries (India, the United States, and South Africa) across ten economic sectors, including Information Technology, Banking, and Pharmaceuticals. Unlike many prior studies that combine numerical and textual inputs, this study relies solely on three years of numerical time series data and examines model responses in terms of decision labels such as buy, sell, or hold. The LLMs were provided with historical closing-price sequences and prompted with three types of finance-related questions: (a) whether to buy a stock, (b) whether to sell or hold a stock, and (c) in a pairwise comparison, which stock to buy or hold. These prompts were evaluated across two investment horizons: 1 month and 3 months. Model outputs were compared against realized market outcomes during the corresponding test periods. Performance was assessed across four key dimensions: country, sector, annualized volatility, and question type. The models were not given any supplementary financial information or instructions on specific analytical methods. The results indicate that GPT-4.0 achieves the highest average accuracy (56%), followed by LLaMA-4-Scout-17B-16E (48%) and Gemini 2.0 Flash (39%). Overall performance remains moderate and varies across market conditions, with relatively higher accuracy observed in high-volatility regimes (51%). This work evaluates how LLMs behave when presented with structured numerical price sequences in a controlled decision-labeling setting and contributes to the broader discussion on the potential and limitations of LLMs for numerical decision tasks in finance.
