A flood of information has occurred, making it challenging for people to find and filter their favorite items. Recommendation systems (RSs) have emerged as a solution to this problem; however, traditional Appenrecommendation systems, including collaborative filtering, and content-based filtering, face significant challenges such as data scalability, data scarcity, and the cold-start problem, all of which require advanced solutions. Therefore, we propose a ranking and enhancing sequence movie recommendation system that utilizes the combination model of deep learning to resolve the existing issues. To mitigate these challenges, we design an RSs model that utilizes user information (age, gender, occupation) to analyze new users and match them with others who have similar preferences. Initially, we construct sequences of user behavior to effectively predict the potential next target movie of users. We then incorporate user information and movie sequence embeddings as input features to reduce the dimensionality, before feeding them into a transformer architecture and multilayer perceptron (MLP). Our model integrates a transformer layer with positional encoding for user behavior sequences and multi-head attention mechanisms to enhance prediction accuracy. Furthermore, the system applies KMeans clustering to movie genre embeddings, grouping similar movies and integrating this clustering information with predicted ratings to ensure diversity in the personalized recommendations for target users. Evaluating our model on two MovieLens datasets (100 Kand 1 M) demonstrated significant improvements, achieving RMSE, MAE, precision, recall, and F1 scores of 1.0756, 0.8741, 0.5516, 0.3260, and 0.4098 for the 100 K dataset, and 0.9927, 0.8007, 0.5838, 0.4723, and 0.5222 for the 1 M dataset, respectively. This approach not only effectively mitigates cold-start and scalability issues but also surpasses baseline techniques in Top-N item recommendations, highlighting its efficacy in the contemporary environment of abundant data.

Aging population is a worldwide trend, which has created a societal crisis as many countries face the challenges of supporting an aging population with increasing costs of healthcare and decreasing numbers of caregivers. Sleep related disorders are common diseases, especially among the elderly. In this paper, we propose a simple and affordable unobtrusive sensing environment including a high-sensitive accelerometer on a bed and passive infrared (PIR) motion sensors in every room, following the generic framework of Internet of Health Things (IoHT) for monitoring the elderly’s sleep-wake conditions, to assess their sleep quality. The environment is nonintrusive, comfortable and can be used for long-term sleep monitoring, detecting early symptoms of sleep related disorders, and responding to caregivers. We implement and pilot test the environment under different daily living situations related to sleep quality. We develop a feature extraction algorithm and applied five popular data analytics models to assess their relative performance. Our study shows that all classifiers except Naïve Bayes can effectively detect sleep quality with the Area under ROC curve (AUC) performance higher than 90%. Among which multilayer feed-forward neural network achieved the best results, in which the detecting sensitivity is up to 96.61%, specificity is 91.81% and AUC performance is 94.21%.

Dementia, a syndrome which is characterized by a decline in cognitive abilities such as memory, thinking, behavior, and the ability to perform daily living activities, is prevalent in people aged 60 and above. However, detecting it early enough can possibly slow its continuous degeneration and lessen the toll on families and caregivers alike. Due to its mortality within 10 years of onset as well as its enormous socioeconomic burden, there have been active efforts by researchers to find smart and innovative solutions for its early detection, prediction, monitoring, and management. These efforts are driven by the recent advancements in the Internet of Things (IoT), wearable technologies, and machine learning algorithms. The solutions are modeled around the modifiable risk factors of dementia. In this study, we conducted a survey of smart solutions developed or implemented to assist caregivers and clinicians in managing the health of these affected individuals. We then looked at the issues and limitations of these solutions, and argued that integrated solutions comprising wearable and non-wearable technologies modeled around multiple risk factors of dementia are necessary and should be the direction of future studies.

Estimating the concentration levels of gas components in a mixture is vital for taking preventive actions and ensuring public safety in various environments, including industrial settings, urban areas, and confined spaces where gas leaks or accumulation can pose significant health and safety risks. This is because general gas sensors for target pollutants are easily affected by other gases. In this study, we assessed the effectiveness of combining data segmentation, feature extraction, and normalization on the performance of three deep learning models: one-dimensional Convolutional Neural Network (CNN-1D), Long Short-term Memory (LSTM), and Gated Recurrent Unit (GRU) network for predicting gas levels in two combinations: ethylene-methane and ethylene-carbon monoxide (CO). We evaluated performance using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-squared (R²). Among the models evaluated, the GRU network outperformed the CNN-1D and LSTM in terms of averaged MAE (0.1578), RMSE (0.0481) and R² (0.96375). Comparative analysis with existing studies confirmed that our proposed GRU model outperformed others so that we propose its use in gas concentration estimation tasks.

In this paper, we propose a respiration monitoring system based on the channel state information received from a commercial off-the-shelf Wi-Fi device. We develop data processing modules for extracting respiration parameters from the Wi-Fi signal and estimate the relative tidal volume. In order to verify the performance of the proposed monitoring system, we use a human patient simulator and also test one real human subject in various experimental settings. With the human patient simulator, we evaluate the monitoring performance at various tidal volume levels and compare our monitoring results with the simulator settings. In testing with a human subject, we test our approach on different sleep postures and positions, Wi-Fi router’s positions, and monitoring environments and compare our performance with the spirometer measurement. The results with both the human patient simulator and real human subject demonstrate the strong and robust estimation performance of the proposed approach in various monitoring conditions, indicating the high promise for non-intrusive respiration monitoring in practice.

Monitoring the curvature of the lumbar spine is important for determining the incidence of lower back pain and other spinal disorders in individuals undergoing physical therapy and rehabilitation and in the field of sports medicine. Especially, to recognize and prevent habitual incorrect spinal curves, a well-suited measurement system is required. In this study, a wearable smart sensing system integrating four flexible sensors and three inertial measurement unit sensors with machine learning was developed. The proposed system was tested on 20 subjects to evaluate its performance. In the experiment, 11 postures were tested using five classes as targets. A feature extraction algorithm was proposed for generating 52 features based on a combination of seven different sensor signals and building classification algorithms for detecting spine events based on the extracted features. The accuracies for classifying five levels of spine curves were 99.38% overall and 99.79% in a tenfold cross-validation test, respectively. The proposed method can estimate spine curve class levels without personalized calibrations.