Sentiment analysis has become an important area of research in natural language processing. This technique has a wide range of applications, such as comprehending user preferences in ecommerce feedback portals, politics, and in governance. However, accurate sentiment analysis requires robust text representation techniques that can convert words into precise vectors that represent the input text. There are two categories of text representation techniques: lexicon-based techniques and machine learning-based techniques. From research, both techniques have limitations. For instance, pre-trained word embeddings, such as Word2Vec, Glove, and bidirectional encoder representations from transformers (BERT), generate vectors by considering word distances, similarities, and occurrences ignoring other aspects such as word sentiment orientation. Aiming at such limitations, this paper presents a sentiment classification model (named LeBERT) combining sentiment lexicon, N-grams, BERT, and CNN. In the model, sentiment lexicon, N-grams, and BERT are used to vectorize words selected from a section of the input text. CNN is used as the deep neural network classifier for feature mapping and giving the output sentiment class. The proposed model is evaluated on three public datasets, namely, Amazon products’ reviews, Imbd movies’ reviews, and Yelp restaurants’ reviews datasets. Accuracy, precision, and F-measure are used as the model performance metrics. The experimental results indicate that the proposed LeBERT model outperforms the existing state-of-the-art models, with a F-measure score of 88.73% in binary sentiment classification.

Kenya’s increasingly evolving power grid requires advanced data stream mining techniques for more effective real-time monitoring and control. This paper examines how these techniques are currently being used in Kenya’s smart power networks, highlighting both their advantages and disadvantages. Although the current systems support basic monitoring and fault detection, they are unable to effectively handle high data velocity, variability, and the integration of renewable energy. High prices, weak regulatory support, limited technical skills, poor data quality, and inadequate computational infrastructure are some of the main obstacles. This paper identifies machine learning, adaptive clustering, and edge computing as possible solutions for improving fault detection, dynamic load balancing, real-time monitoring, and the integration of renewable energy. These strategies optimize grid performance while addressing latency and scalability challenges. The study recommends implementing advanced data stream mining techniques, making significant infrastructural investments, boosting technical capacity, and creating supportive legislative frameworks in order to improve Kenya’s smart power grid. Implementing these changes may help Kenya to enhance its grid resilience and reliability, boost energy efficiency, and promote long-term sustainability. Clinical trial registration: Not applicable.

The Internet has become a vital resource for health-related information, advice, and support, with more people turning to websites, social media, and online groups for guidance on various health topics ranging from healthy living to disease diagnosis and treatment. The rise of smart devices connected through the Internet of Things (IoT) has further expanded access to health services, enabling Health Recommender Systems (HRS) to collect patient data and generate personalized recommendations. Despite its potential, the overwhelming volume of health information available through the Internet, smart devices, and connected technologies often hinders users from efficiently accessing timely, relevant, and personalized insights, thereby limiting their ability to fully benefit from these resources. Recommender Systems (RS), widely used in e-commerce and social networks to help users navigate choice overload, have also been adapted to healthcare and the Internet of Things (IoT), where they assist patients, medical professionals, and smart devices in delivering personalized guidance, automation, and decision support. Context-aware collaborative filtering presents a promising solution, allowing recommendations to be dynamically adapted to users' specific circumstances. Integrating contextual information in HRS aim to enhance predictive accuracy and improve the overall quality of the recommendations. Context-aware health recommender systems (CAHRS) incorporate contextual information to dynamically tailor recommendations to individual circumstances. This integration enhances predictive accuracy and significantly improves the overall quality and relevance of health-related suggestions. However, the absence of a standardized framework to design, develop, implement, evaluate and deploy HRS and CAHRS limits their effectiveness and scalability. The paper outlines, evaluates and validates a framework for Context-aware collaborative filtering Recommender System for Health (CARSH) using Differential Context Weighting (DCW) and Particle Swarm Optimization (PSO) and a proposed hybrid context matching similarity measure of Adjusted Triangle multiplying weighted Jaccard (ATMWJ). Their performance is compared with context-aware modeling approaches that rely on weighted matching context similarity. The experimental results show that the proposed hybrid context similarity measure significantly outperforms existing approaches with MAE of 0.5325 and RMSE of 0.7371 compared to the baseline of 0.5901 and 0.7802 respectively on LDOS-CoMoDa dataset. On the Drug Reviews dataset, ATMWJ achieved an F-measure of 0.9239, surpassing the 0.65 baseline and confirming its superior recommendation quality. The experimental results demonstrate that this approach significantly improves prediction accuracy over existing methods, highlighting the importance of selecting an appropriate similarity metric for enhancing recommendation relevance.