This review article examines different architectural approaches for IoTbased Remote Patient Monitoring Systems. Four different classification categories have been identified, each with their own strengths and weaknesses. Fog-based architectures reduce latency and enhance privacy through localized data processing. Cloud-based solutions offer robust storage and analytics capabilities, but struggle with latency and centralization. Hybrid systems combine edge and cloud computing to improve scalability and efficiency, while blockchain-enabled architectures focus on secure, decentralized data management. The findings from 11 original publications were summarized. This paper describes the characteristics of the identified approaches by examining common and unique concepts used by the solutions in the same classification category based on prior analysis.

nema

The accelerated pace of digital transformation has significantly reshaped the cybersecurity domain, fostering an interconnected ecosystem in which cyber threats have expanded in both their complexity and scope. Traditional cybersecurity methods are increasingly inadequate for addressing the rapidly evolving threat landscape, emphasizing the critical need for intelligent, adaptive, and proactive defensive strategies. This study introduces Dynamic Industrial Cyber Risk Modelling Based on Evidence (DICYME), a comprehensive system that integrates diverse analytical techniques to identify patterns and characteristics that reveal emerging threat trends, enabling organizations to proactively defend against potential future attacks. Beyond threat detection, DICYME operates as a pipeline that retrieves data from diverse cyber incident reports, specialized databases, and other relevant sources of cyber-related information, applies specialized techniques for victim identification, indicator computation, threat actor profiling, Common Vulnerability and Exposure (CVE) relationship mapping, and ultimately performs the Cyber Risk Quantification (CRQ). This final stage represents the system?s most distinctive contribution, as it translates complex analytical outputs into actionable risk insights, empowering organizations to make informed strategic decisions in the face of evolving cyber threats. Alternatively, the system implements an automatic workflow that constructs new datasets of compromised entities, enabling these datasets to be used by all components of the system. Experiments on real cyber incident datasets demonstrate the system?s ability to automatically construct high-quality victim profiles and estimate annualized financial risk, offering a scalable and data-driven approach for proactive cybersecurity management.

To improve the mental health education level of college students and promote the adoption of artificial intelligence (AI) in college education, freshmen from a university are selected as research subjects. Two classes are chosen, with 40 students in each, and they are assigned as the experimental and control groups, respectively. The analysis of the voice interaction system of campus psychological consultation is combined with self-efficacy. The Sixteen Personality Factor Questionnaire (16PF) and the General Self-Efficacy Scale are adopted to analyze the mental health of the two groups of students at the beginning and the end of the semester. The interactive technology in AI is applied to construct a user mental model to achieve voice interaction designing. Knowledge base matching is performed on students' interactive input text, and long short-term memory (LSTM) is adopted to analyze the sentiment type of the input information and then classify it. Moreover, the answer with higher voice confidence is returned as the candidate's answer to the machine. After that, the possible text of each candidate answer is predicted and analyzed according to the interactive condition probability, and the optimal result is fed back to the student interface. Then, it points out the effect of the proposed method in analyzing the influence of inspirational film appreciation courses on the mental health level of college students. The results show a significant difference in the sensitivity factor in the experiment results at the beginning of the semester (P<0.05). Meanwhile, the difference in the automaticity factor is very significant (P<0.05). Both sensitivity and automaticity factors show different variations than expected in the experimental data at the beginning of the semester, especially the changes in automaticity factors. At the end of the semester, there are no significant differences in the sensitivity and self-discipline of the students. Moreover, it significantly influences students in the experimental group after enjoying the inspirational film, especially in terms of interpersonal communication and emotional management, positively impacting students' mental health. Meanwhile, the students in the experimental group adopt the voice interactive system for mental health consultation, and the machine can give some references to protect students' privacy to a certain extent. Therefore, when adopted to analyze the impact of inspirational film appreciation courses on the mental health of college students, the voice interaction system for campus psychological consultation under AI combined with self-efficacy has a positive effect. It also contributes to recommending college mental health education and the adoption of AI in college education.

Malicious websites pose significant social risks, necessitating automatic, efficient, and accurate identification methods. This paper proposes a POST traffic classification method based on website templates to identify abnormal traffic from gambling websites. Using Fiddler, POST message data is collected from several gambling sites, extracting features like URLs, cookie parameters, and request body parameters to create a Gambling Website Single POST Message Dataset (GSPD). These features are converted into vector representations withWord2Vec and TF-IDF techniques. Hierarchical clustering identifies template-generated types, achieving unsupervised template recognition. Using clustering results, individual POST messages are labeled and features are extracted using TF-IDF and mutual information methods. The parameters of a Support Vector Machine (SVM) are then optimized with the Particle Swarm Optimization (PSO) algorithm for optimal classification. Experimental results show the model?s excellent performance, with a test set accuracy of 0.9985 and high precision, recall, and F1-scores, effectively identifying gambling and other illegal websites.

Explainable Artificial intelligence (XAI) represents a pivotal innovation aimed at addressing the ?black box? problem in AI, thereby enhancing users? understanding of AI reasoning processes and outcomes. The implementation of XAI is not merely a technological endeavor but also involves various individual factors. As XAI remains in its early developmental stages and exhibits unique characteristics, identifying and understanding the factors influencing users? intention to adopt XAI is essential for its long-term success. This study develops a research model grounded in the characteristics of XAI and prior technology acceptance studies that consider individual factors. The model was evaluated using data collected from 252 potential XAI users. The validated model exhibits strong explanatory power, accounting for 45% of the variance in users? intention to use XAI. Findings indicate that perceived value and perceived need are key determinants of users' intention to adopt XAI. These results provide empirical evidence and deepen the understanding of user perceptions and intentions regarding XAI adoption.

This work aims to optimize college football training using deep learning techniques, addressing the inefficiencies, difficulty in action recognition, and insufficient data analysis present in current training methods. An intelligent optimization system combining Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) is proposed to tackle these challenges. Compared to traditional single models, the Convolutional Neural Network-Recurrent Neural Network (CNN-RNN) architecture remarkably improves the efficiency and accuracy of processing training data by leveraging the strengths o spatial features and temporal sequence features. The experimental results show that CNN-RNN model is significantly superior to the traditional 3D CNN model and other advanced models, such as Transformer, Long Short-Term Memory (LSTM), Bidirectional LSTM and Gated Recurrent Unit (GRU), in key indicators such as accuracy, precision, recall and F1 score. Specifically, CNN-RNN model achieves 92.5% accuracy, 91.2% precision, 93.1% recall and 92.1% F1 score. The lowest training loss rate is 0.24, which is significantly better than other models. In addition, the introduced data balance strategy effectively improves the prediction performance of a few categories (such as foul and yellow card events) through oversampling, undersampling and weighted loss function, and further enhances the generalization ability and practicability of the model. Future research focuses on expanding the dataset, further improving the model?s generalization ability, and exploring its application in real training scenarios.

Class imbalance is a common problem in machine learning where the majority class has a significantly higher number of instances than the minority class, which leads to bias towards the majority class. The problem can be effectively addressed by using Generative Adversarial Network (GAN) to generate realistic synthetic samples. In this work, we present a GAN-based approach that makes use of hybrid models that combine oversampling techniques with undersampling and ensemble techniques to reduce overfitting. The proposed approach was evaluated on two datasets with different level of class imbalance using six widely used classifiers and compared with two popular class balancing techniques ? SMOTEENN and SMOTETomek. The results show that the proposed approach outperforms them in highly imbalanced datasets.

Typically, the image features are compared to find the similarity among the images in a content-based image clustering system. However, images with high feature similarity may be different from each other in terms of semantics. Hence, this paper proposes a novel algorithm based on unsupervised neural classifier systems for in-vivo image clustering to address the semantic gap issue. The visual features are represented using Wavelet transform and Zernike moments, and a selforganizing map is utilized for the clustering of images. The algorithm-based prototype system is trained for categorizing gastral images in the respective clusters as per the similarity. The system can be used to segment images with automatic noise reduction and rotation invariances for given images. Experiments are performed on the real gastrointestinal images obtained from a known gastroenterologist, and the results using Daubechies Wavelet Transform + Zernike Moments on LUV color scheme yield 88.3% accuracy.

Programming languages provide the notation for writing computer programs capable of granting our devices the desired functionalities. Even though they may seem intangible, the resulting programs involve an amount of energy consumption, which has an impact on the environment. Some studies on the consumption of programming languages indicate that while being one of the most widely used languages, Python is also one of the most demanding in terms of energy consumption. To provide developers using Python with a set of best practices on how to use it in the most energy-efficient way, this paper presents a study on whether the different ways of programming in Python have an impact on the energy consumption of the resulting programs. We have studied the relationship between Python?s energy consumption and the fact that Python is a very versatile language that allows programs to be compiled and executed in many different ways. From the results obtained in our study, there seems to be a clear relationship between software energy consumption at runtime and: (1) the use of interpreted or compiled Python; (2) the use of dynamically or statically typed variables. Compiling Python code is a good option if it is done using the py compile module. The use of interpreted code seems to decrease energy consumption over compiling using Nuitka. The use of dynamically typed variables seems to decrease considerably the graphics and processor energy consumption. In addition, we have observed that energy consumption is not always directly related to execution time. Sometimes, more power in less time increases consumption, due to the power required.