Community engagement is essential to social service delivery, yet traditional community needs assessment remains time-consuming and poorly suited for timely monitoring. This study proposes a semi-supervised learning framework to identify emerging community needs and service gaps from massive, mostly unlabeled, unstructured text. We construct an explicit heterogeneous text graph where each record is a document node linked to keyword and need-category nodes; document–document edges are built using a weighted combination of semantic similarity (BERT cosine), lexical overlap (keyword Jaccard), and temporal proximity. A graph neural network with iterative self-training leverages 3% expert-labeled seed data and the remaining unlabeled corpus to classify records into a 10-category need taxonomy. On 176,602 records, the proposed model achieves F1 = 0.895 and Recall = 0.899, outperforming supervised baselines trained on the same labeled ratio by 23.8% (macro-F1). Post-hoc quarterly aggregation of predictions enables trend monitoring and prioritization of service-gap severity for decision support.

The rapid rise of mobile technology paired with the steady growth of the internet, has led to a massive increase in the amount of user generated content, such as online consumer reviews, accessible through the browser. As the volume of user-generated content continues to rise, it becomes increasingly important to develop sophisticated methods for performing sentiment analysis on the texts collected from users, especially those that have been generated in relation to restaurants and similar types of service establishments. In this paper, we will present a new approach to sentiment analysis which incorporates Latent Dirichlet Allocation topic models, Term Frequency- Inverse Document Frequency vector representations and XGBoost Classifiers into a unified framework. Unlike conventional implementations, this study integrates probabilistic topic distributions from LDA with multi-level n-gram TF-IDF features and evaluates their combined impact using XGBoost for enhanced classification performance. Using three distinct n-gram levels (unigrams, bigrams, and trigrams), we will evaluate various aspects of text-based data including common linguistic patterns and sentiment trends. Higher-order n-grams were included to capture contextual dependencies beyond single-word features. Overall, our results demonstrate that the performance of our proposed framework is superior to traditional corpus-based models on multiple evaluation metrics, including: classification accuracy 96.07%, classification sensitivity 95.43%, classification specificity 97.12% and F1-Score 96.16%. 

Forecasting time series data especially in volatile sectors like financial markets, shows significant challenges due to non-linearity, non-stationarity and noise in the data. Traditional forecasting models most likely fail to generalize effectively across varying tasks without extensive retraining. This study investigates the application of meta learning techniques, particularly First-Order Model-Agnostic Meta-Learning (FOMAML) and Reptile, to make adaptability and generalization better in time series forecasting tasks. An extensive empirical study was done using three neural networks as base models, namely Long Short Term Memory (LSTM), Gated Recurrent Unit (GRU) and Feed Forward Neural Network (FFNN) applied to four real-world stocks: TCS, TATASTEEL, GRASIM and DJIAHD. The models were evaluated under few-shot learning(defined here as 211-shot learning using sliding window samples) conditions with varying iteration counts(outer loops or epochs) and their effectiveness was checked using some common standard metrics like RMSE(Root Mean Squared Error), MAE(Mean Absolute Error) and R²(Coefficient of Determination). Outcomes have shown that meta-learning approach notably performs much better than traditional models with MAML(First Order) in particular showing quicker task adaptation as well as stable convergence behavior, especially when it used with GRU and LSTM as base models, as validated empirically on the GRASIM dataset where the MAML with LSTM configuration attained around 81.9\% reduction in RMSE (dropping the value from 622.94 to 112.60 over the iterations). In all four stocks, reptile shows relatively steady performance. The study validates the potential of meta-learning as a powerful framework for time series forecasting problem in dynamic settings which offers robust algorithmic foundation for numerous future financial modeling applications.

It is well known that diagnosing Alzheimer's disease (AD) accurately and early is a major clinical challenge, especially when using brain MRI data to differentiate between subtle stages of cognitive decline. This study investigated the efficacy of two deep learning models for the classification of AD stages: Vision Transformer (ViT), a transformer-based architecture, and EfficientNetB7, a convolutional neural network. To enhance classification performance and address class imbalance, extensive data preprocessing and augmentation techniques were employed on the publicly accessible 'Alzheimer’s Dataset (4 class of Images)' from Kaggle. This dataset comprises 6,400 brain MRI images categorized into four AD stages: Non-Demented, Very Mild Demented, Mild Demented, and Moderate Demented. Techniques applied included cropping, horizontal and vertical flipping, 20-degree rotations, histogram equalization, Gaussian noise addition, Gaussian blurring, and thresholding, aimed at improving the representation of underrepresented classes. Hyperparameter optimization was executed via a two-phase methodology: an initial grid search to determine parameter ranges, succeeded by Bayesian optimization employing an upper confidence bound acquisition function to refine learning rates, batch sizes, momentum, and weight decay values. Experimental results indicated that EfficientNetB7 attained a classification accuracy of 93.5% with F1-scores surpassing 92% for early-stage classes, whereas Vision Transformer (ViT) recorded a lower accuracy of 88.7% and exhibited diminished sensitivity to early-stage instances. The performance disparity is due to ViT's dependence on extensive training datasets, which may restrict its generalization when utilized on comparatively smaller medical imaging datasets. The results indicate that, in dataset-constrained  
scenarios, CNN-based architectures such as EfficientNetB7 may provide more consistent and effective performance. Using distinct training, validation, and test datasets, the model's generalization, training stability, and computational efficiency were assessed. With an intuitive user interface, the top-performing model, EfficientNetB7, was implemented as a web-based application to facilitate real-time supportive predictions for research demonstration. This comparative analysis demonstrated that the CNN-based EfficientNetB7 exhibited more robustness with constrained medical imaging data and was computationally economical, but the transformer-based ViT displayed increased sensitivity to dataset size and necessitated extended training to attain similar convergence. The development of a validated and deployable AI-based Alzheimer's disease diagnostic solution showed great promise for clinical use.

Advances in deep learning have highlighted the need for models tailored for deployment in resource-constrained environments (RCEs), where memory and processing limitations present significant challenges, such as those found in mobile devices, Internet of things (IOT) devices, and embedded systems. This paper introduces GRMobiNet, a novel deep neural network (DNN) model designed to address these challenges in image classification tasks by balancing computational complexity with model accuracy in RCE settings. The model focuses on key performance goals inspired by previous state-of-the-art models, aiming to achieve a better balance between complexity and accuracy. These goals include reducing the model's computational complexity to fewer than 4 million parameters, limiting memory usage to under 16 megabytes, and achieving an accuracy greater than 80%. By meeting these objectives, GRMobiNet enhances both the effectiveness and efficiency of deep neural network deployment in RCE settings. GRMobiNet builds upon MobileNet as its baseline, incorporating advanced techniques such as depthwise separable convolutions, compound scaling, global average pooling, and quantization to optimize performance. Trained on ImageNet-10, a subset of ImageNet-1K, the model underwent rigorous performance evaluation. Experimental results demonstrate that GRMobiNet achieves its performance objectives, with a computational complexity of 3.2 million parameters, memory utilization of 12.6 megabytes, and a prediction accuracy of 92%, validating its suitability for RCEs. This research presents a scalable framework for balancing accuracy and computational efficiency, with significant implications for RCE devices. In future work, GRMobiNet will be tested on commercially available RCE mobile devices using real-world images to assess its practicality and evaluate its performance in terms of accuracy, confidence, and inference time for image classification in real-world scenarios.

Public transport (PT) users often experience instances of leaving items behind in the public transport system. Finders who come across these items may choose to keep them maliciously or, out of goodwill, decide to return them. This paper aims to utilize six (6) machine learning models, including LR, SVM, DT, RF, NB, and KNN, to predict the ability of finders to return found items. Nine (9) features, comprising four (4) demographic parameters (age, gender, income, and education), were used in the models’ prediction process. The study involved a total of 603 PT users in the Accra cosmopolitan area of Ghana to assess finder’s decision regarding returning found item(s). The classification success rates were obtained as follows: 86.740% (LR), 87.293% (SVM), 82.873% (DT), 85.083% (RF), 85.083% (GNB), and 87.845% (KNN) using Python codes. The RF model also performed well, considering the balance of performance with the desired precision and recall. RF, GNB, and LR achieved the highest AUC values (0.78), demonstrating strong discriminative ability in predicting user honesty. 

Internet addition is a type of mental disorder. It is the result of the excessive internet usage and is concern for the physical and psychological well-being. This paper employs machine learning techniques to understand, evaluate and predict the severity of internet addiction and its impact on health. For this purpose, a real dataset of “Internet Addiction and Mental Health among College Students in Malawi” has been considered. It consists of self-assessed response of 984 university student participants. That includes demographic, behavioral and health-related information. Based on this dataset, two type of relationship have been discussed (1) relation between “demographic features” and “health complexities” and (2) relation between “Internet usage behaviors” and “health complexity”. Next, the key features were identified through comprehensive data analysis. Additionally, there machine learning algorithms namely Backpropagation Neural Network, Random Forest, and C4.5 Decision Tree— were tested to identify ‘internet addiction’ in a subject with the four severity levels (0 to 3). According to results, the Random Forest classifier achieved the highest accuracy of 91%. Additionally, C4.5 algorithm has been used for extracting rules for predicting “Internet addiction” severity level. These rules are demonstrating a relation between “Internet usages pattern” and “Internet addiction severity level”. Additionally, these rules are easy to interpret and can be utilized as a practical tool for self-assessment towards Internet addiction and additionally beneficial for healthcare professionals.

Electronic devices and internet purchasing are more common today. For online line shopping most of people are using internet banking and credit for doing payment for purchasing. For time saving and various offers on credit card and debit card customer prefer on line shopping like various platform Amazon, Flip cart, big basket etc. For online transaction security is prime concern. There is various type of attack possible during online transaction, stealing of password, fraud transaction, and meet in middle attack etc. During the online transaction stealing confidential information like OTP, transfer money from someone account to another account is a crime.   In the digital world fraud during the online transaction day by day increases exponentially. To detect the unauthenticated transaction and fraud during online used various methods. Data is playing very important role during the online fraud. So, knowledge discovery is most frequently used to protect online fraud. In this paper suggested a technique based on knowledge discovery and machine learning methods, we strive to develop the best model possible in this research study to predict transactions involving fraud and transactions involving no fraud. Fraud detection uses a variety of machine learning techniques, including K-Means clustering methods, Support Vector Classifier, Logistic Regression, and Anomaly Detection Algorithm Techniques. After analysis it was found that Anomaly Detection Algorithm Techniques gives best accuracy for fraud detection 99.85%.

Accurate and immediate incident identification is essential in the cybersecurity area, as it allows the timely detection of threats, along with countermeasures and mitigation, ensuring security for organizations and individuals. This reduces false positives and enables efforts to be concentrated on real risks. This paper presents a framework that integrates ontologies and Large Language Models (LLMs) to identify incidents from events within the context of security threats. Ontology rules are employed to infer probable incidents, resulting in an initial set of incidents for analysis. Furthermore, ontologies provide contextual information, which is combined with event data to formulate queries for LLMs. These interactions with LLMs produce a second set of probable incidents. The outputs from ontol-ogy-based inferences and LLM-driven responses are then compared, and the discrepancies are leveraged to refine ontology rules and adjust LLM responses. Experimental results, focusing on context generation and incident detection, demonstrate that the integration of ontologies and LLMs significantly enhances the accuracy of incident identification when compared to using only LLMs.

Concept drift is a critical challenge in dynamic environments, where evolving data distributions can abruptly reduce predictive accuracy. Sudden drift requires reliable detection methods that minimize latency and false alarms, yet traditional detectors often depend on labeled data, delaying adaptation and limiting robustness.
This article introduces Neutrosophic Pseudo Labeling Sudden Drift Detection (N PSDD), a novel framework for unsupervised sudden drift detection based on neutrosophic theory. The method integrates neutrosophic clustering for pseudo labeling, block wise neural modeling, drift quantification via neutrosophic mean deviation, and adaptive threshold evaluation. By explicitly modeling truth, indeterminacy, and falsity, N PSDD captures uncertainty regions that conventional probabilistic measures fail to represent.
Experimental validation on synthetic and real world datasets demonstrates that N PSDD achieves competitive dtection latency (MTTD ≈ 23–35 instances), a lower false alarm rate (FAR ≤ 3.1%), a reduced missing drift rate (MDR ≤ 2.5%), and consistently higher G mean values (up to 0.91) than benchmark methods do. For example, on the Poker Hand dataset, N PSDD achieved MCC = 0.846 and accuracy ≈90%, while on electricity it reached MCC = 0.623 with FAR = 3.1%. In contrast, unsupervised baselines (KS WIN, HDD, MMD) yielded higher FAR (≈6–10%) and lower MCC (≤0.56), confirming their limitations in capturing real concept drift.
Overall, the N PSDD enhances the resilience of learning models under non stationary conditions and provides a robust solution for real time applications, including financial forecasting, fraud detection, and adaptive control systems.

Automatically classifying abstract sentences into significant categories such as - background, methods, objective, result, and conclusions - is an essential support tool for scientific medical database querying that assists in searching and summarizing relevant literature works and writing new abstracts. This paper presents a memory-efficient deep learning model for sentence role classification in medical scientific abstracts, achieved by integrating quantized Transformer layers with a Bidirectional Long Short-Term Memory (BiLSTM) network. While the core components are recognized, our contribution is demonstrated in the successful application of quantization to this hybrid architecture, significantly reducing model size (from ~75MB to ~25MB) without a meaningful drop in classification performance on a subset of the PubMed 200k RCT dataset. This makes our approach distinctively practical for deployment in resource-constrained environments, offering an effective tool for automated literature analysis.

This paper introduces a novel reference model for intelligent longitudinal vehicle control, designed to enhance both safety and passenger comfort. The proposed model dynamically adjusts the follower vehicle’s acceleration based on its penetration distance relative to the lead vehicle, ensuring smooth speed transitions and adaptive deceleration. By preventing abrupt braking, the model maintains a safe inter-vehicle distance while reducing passenger discomfort. Key contributions include an analytical derivation of the follower vehicle’s dynamics and a novel formulation of the safety distance using the Lambert W function, enabling precise parameter optimization. A dedicated optimization framework ensures compliance with safety constraints while minimizing excessive acceleration and jerk. The model’s performance is validated through numerical simulations in various driving scenarios, including emergency braking, steady-speed following, variable-speed adaptation, and stop-and-go traffic. Results demonstrate its effectiveness in maintaining safety while enhancing ride comfort through gradual and controlled deceleration. The proposed approach is computationally efficient and well-suited for Advanced Driver Assistance Systems (ADAS) and autonomous vehicles. Future research will explore its integration with lateral control strategies, real-time adaptability, and machine learning techniques for further performance optimization in dynamic driving environments.