Fusion: Practice and Applications

Melanoma is one of the forms of skin cancer that affects people worldwide. Research indicates that nearly 75% of the global population has been impacted by melanoma. Early detection and treatment of melanoma significantly increase survival rates. However, detecting melanoma in its early stages can be challenging because dermatologists typically rely on visual examination and biopsy analysis, which is both time-consuming and labor-intensive. This highlights the need for automated, efficient methods to identify melanoma at earlier stages. Skin cancer is generally classified into two categories: melanoma and benign tumors. The goal of this study is to facilitate the early detection of melanoma by employing deep learning techniques, specifically convolutional neural networks (CNNs), to distinguish between melanoma and benign lesions using the ISIC dataset. The proposed model achieves an accuracy of 80.80%, outperforming previous approaches by offering faster and more accurate melanoma detection.

The emerging field of healthcare has taken severe measures to safeguard sensitive patient health-related information especially the information taken from the predictive model. In this study, a novel blockchain-based solution is proposed in correlation with the Fuzzy-enhanced CLSTM model (FCLSTM) for storing and transmitting the data securely for heart disease prediction systems by ensuring data integrity, confidentiality, and access control. The proposed model uses a blockchain-based network which is implemented to prevent the tampering or unauthorized access to patients’ health-related data. The process begins with techniques that incorporate the predicted heart disease information from the patient’s data and is encrypted by using the hashing algorithm. A secure hybrid blockchain-based data management framework (SHB-DMF) is designed for exchanging the patient’s health data which enhances scalability and accessibility to the healthcare environment. The system incorporates a SHAES-256 hybrid model for enhancing the data confidentiality and integrity before transmitting to the neural network (FCLSTM). The proposed model uses a smart contract for regulating data access by ensuring the entry of the authorized entities by providing a suitable decrypting mechanism and interacting with the patient’s data. The smart contracts can automate the data retrieval workflows by integrating the blockchain seamlessly with the prediction model. The security process is a three-phase process that includes defining the nodes, selecting of consensus mechanism, and establishing the governance structure for facilitating secure operations. The security and load testing ensure resilience to potential cyber threats and the scalability required for handling high transaction volumes of medical data. Deploying the proposed system provides a robust infrastructure that is tamper-resistant thus advancing the reliability of the cardiovascular prediction system.

Stock price forecasting has oftentimes interested several researchers around the world. Making predictions for the future largely depends on the data that will be used to train the model. In general, historical data are used to train models, which contain a features of different types, out of which, not all are necessarily helpful in making predictions. It is, hence, crucial to select the features that can be most useful to make precise predictions. This article proposes a feature selection approach based on the K-means clustering algorithm and elastic net regularization. We have used the K-means algorithm to cluster all the correlated features together and apply elastic net regularization to select the most predictive features within each cluster. We use the selected features to train an LSTM model which predicts the future closing price of a stock for the upcoming trading day. We evaluate the performance of our proposed approach in comparison to the existing approach and observe performance improvement.

With the latest developments in computer vision, processing, accurate deepfakes (DF) require powerful tools. Recent research has developed a useful technique for identifying DFs in networks. The inter-frame differences of the gathered media streams, however, are beyond the scope of most methods. In this research, an Optimized Gaussian Convolutional Neural Network Framework for Enhanced Detection of Deepfakes in Digital Media (OGCNN-DDF-DM) is proposed. Initially the input images are gathered using the Face Forensics++ (FF++), and Deep Fake Detection Challenge dataset (DFDC) datasets. Then the Multi-Window Savitzky-Golay Filter (MWSGF) is used to improve quality of the DF images and reduce noise. Afterwards, Simple Contrastive Graph Clustering (SCGC) achieves segmentation. Here, the image's facial regions are segmented. Then, the texture features are extracted using Revised Tunable Q-Factor Wavelet Transform (RTQWT) is introduced. The extracted features are fed to Gaussian Convolutional Neural Network (GCNN) to categorize the image as real or fake. Finally, Gooseneck Barnacle Optimization Algorithm (GBOA) is proposed to improve the GCNN classifier. Performance parameters including accuracy, precision, recall, specificity, ROC, and computation time are examined. The introduced method attained an accuracy of 99.6% and the precision of 98.9% on the FaceForensics++ dataset, and 99.5% and 98.6% on the DFDC dataset, respectively.

Steganography conceals ”secrets” within an convenient and expedient multimedia carrier. The carrier could be text (i.e., not plain text), images, audio and/or video files (i.e., carrier channels). The fact that concealed information is contained in the otherwise ordinary and mundane carrier file is known only by the sender-receiver pair. Only they share the existence of the secret. Images are the most popular (i.e., multimedia) carriers because of their inherent property that enables better obfuscation. Content adaptive image steganography is a new trend in the field for messaging secrets inside unsuspected image file transfers. As the name suggests, the embedding locations are altered adaptively depending on the image content that optimizes the decision of choosing a location inside the carrier so that an embedding is not discernible (i.e., additive distortion is minimized). Herein, we critique the various approaches used for content-adaptive image steganography which can be broadly categorized as CNNbased, GAN-based, along with minimizing additive distortion function-based. We provide a brief historical account toward better anticipating the future research opportunities in terms of properties, and evaluation metrics. A summary table of these past and future directions is provided. Moreover, we highlight trends along with their concomitant advantages and disadvantages toward identifying opportunity gaps.

The presented methodology provides an innovative way to answer a question that is rarely observed in academic literature: How can complex data issues like multiple class imbalance be solved using the available models in a simple and efficient way? In this approach, observations are modeled without additional preprocessing. Several classification models including Random Forest (RF), Support Vector Machines (SVM), and Decision Tree (DT) are utilized for conducting the classification analysis. The parameters of these models and the cross-validation function are adjusted to each individual set of observations. This approach has not been researched in depth. We test it about class imbalance in the target variable. Our results demonstrate the benefits of the proposed method.  First, parameter tuning of ML models can be an effective strategy to handle class imbalance. Second, random shuffling prior to cross validation can be a key to resolving the bias coming from multiclass imbalance. Another important finding is that the best results can be achieved when random shuffling, cross validation and parameter tuning are combined. These findings are key to handling class imbalance in classification. Therefore, this research extends the opportunities to handle class imbalance in a simple, quick, and effective way in cases without adding additional complexity to the model.

For data security and integrity, the sharing of Electronic Health Records (EHRs) utilizing blockchain is becoming a vital vision. However, blockchain and storage wielded in prevailing studies arises security and scalability issues. To overcome these issues, this paper proposes a novel Quadratic Interpolation-based Brownian Motion-Double Elliptic Curve Cryptography (QI-BM-DECC)-centric EHR securing in Hyper-Ledger Blockchain (HLB) with Inter-Planetary File System (IPFS). Primarily, the patient and doctor are registered on the hospital website; then, the keys and QR codes are generated for the patient. After that, the patient login with the credential details, QR code, and the purpose of login. The patient did the online consultation booking after successful login; then, the consultation is done grounded on the time scheduled by the doctor. Afterward, the patient securely uploads the EHR on the HLB with IPFS utilizing QI-BM-DECC. Meanwhile, an attribute-centric hashed access policy is created with the selected attributes. After that, utilizing the Mean Public keys- Digital Signature Algorithm (MP-DSA) approach, the hashed access policy is signed. When a doctor request for EHR access, the signature is verified and the access request is sent to the patient. Now, the doctor downloads the EHR from IPFS after being accepted by the patient. The experiential outcomes exhibited the proposed technique’s dominance over the other mechanisms.

Cooperative Intelligent Transportation Systems (C-ITS) cannot work effectively if they do not have both efficient traffic management and solid security. We put forward in this paper an original framework that takes advantage of the Next Generation Simulation (NGSIM) dataset to improve traffic flow and system security by identifying False Data Injection Attacks (FDIA). By applying leading machine learning algorithms to authentic traffic data, we generate models that support improved vehicle coordination as well as provide assistance with security vulnerabilities in C-ITS systems. We are concentrating our method on the optimization of traffic dynamics by making intelligent decisions, while keeping the system secure from malicious cyber attacks. Analyses of the NGSIM data revealed that our proposed approaches produced important advancements in traffic flow efficiency and the accuracy of anomaly detection. Results prove that our framework minimizes congestion and concurrently enhances the reliability and security of collaborative vehicle systems. This investigation proposes a practical approach for fusing traffic optimization with cybersecurity, improving smart city evolution and the future of autonomous vehicles and vehicle connectivity.

Organizations use fusion data modeling to integrate multiple data sources and build precise representations that achieve better organizational clarity. One recent method that has proven effective in many benchmark tests is the arithmetic optimization algorithm (AOA). AOA applies basic distribution behavior to arithmetic operations such as multiplication, division, addition, and subtraction. This paper focuses on the innovative application of AOA in addressing the feature selection problem. The binary version of this algorithm (BAOA) is introduced to solve problems of binary nature. The main part of this version is the transfer function that converts a continuous search space into a discrete search space. Therefore, a new Fountain-shaped transfer function is proposed to enhance global exploration and local exploitation in the BAOA algorithm. The performance of the proposed Fountain-shaped transfer function has been compared with V-shaped and S-shaped transfer functions. Based on ten public datasets, the performance of the proposed transfer function is validated. The Experimental results show the superiority of the proposed Fountain-shaped transfer function not only in getting high classification accuracy with few selected features but also requires inexpensive computational costs.

Presently, smart sensors ensure commercial decisions where integrated electronic systems can be securely organized using blockchain and quantum computing because of their unique characteristics and features. In the current scenario, large-scale quantum computers can be built in which most current cryptographic systems can be hacked. Since digital and quantum computers can conduct computations simultaneously, a quantum tool for blockchain framework design is required. Based on these concerns in this research, an enhanced quantum-assisted blockchain security model using the artificial intelligence (EQ-BSM-AI) technique has been proposed. This model validates cryptosystems and blockchain technologies to determine their vulnerability to quantum attacks. Further, in this model, quantum assisted edge computing technique has been used to model the Human-centric Internet of Things (HIoT) system by introducing a quantum key generation process. Based on the post-quantum blockchain (PQB), a secured cryptosystem that is highly resistant to quantum computer attacks has been introduced in this research. This quantum channel with multiple inputs and outputs (MIMO) is designed for a quantum-based communication system to make this model more efficient and withstand errors. In EQ-BSM-AI, an improved quantum encryption algorithm (IQEA) stores the keys for encryption with a generalized probability accumulation model. For the current quantum computers and communications, our proposed system resulted in an improved sampling error reduction of 12.4%, enhanced efficiency of quantum entanglement of 96.3%, information randomness of 93.9%, correlation analysis of 93.2%, and increased resistance to quantum computing attacks of 90.8% when compared with other existing approaches.

Melanoma is one of the most aggressive types of skin cancer, and its early detection is critical to improving survival rates and treatment outcomes for patients. Conventional diagnostic methods often suffer from high computational costs and low accuracy, primarily due to inadequate feature selection and classification strategies. The goal of this research is to combine state-of-the-art deep learning techniques with optimization algorithms to develop a precise and efficient predictive system for melanoma detection. In this work, we propose a novel framework that integrates Convolutional Neural Networks (CNNs) for image classification and a binary Grey Wolf Optimization (GWO) algorithm for feature selection. The binary GWO algorithm identifies the most relevant features from dermatological images, eliminating redundancy and reducing the computational burden. The CNN is then trained on the refined feature subset to enhance classification efficiency. Extensive experiments on publicly available skin lesion datasets demonstrate that the proposed model significantly outperforms traditional machine learning models. Improvements in sensitivity, specificity, and overall classification accuracy highlight the effectiveness of combining deep learning with optimization techniques. Our results show that deep learning and optimization methods, such as the binary GWO algorithm, can be successfully applied to melanoma diagnosis. This strategy not only improves detection efficiency and accuracy but also supports early diagnosis and treatment planning, leading to better patient outcomes. By leveraging the binary GWO algorithm to optimize the feature selection process and CNNs for image classification, the proposed approach reduces computational costs while increasing classification accuracy. When trained and evaluated on publicly available skin lesion datasets, the model demonstrates significant improvements in sensitivity, specificity, and overall accuracy compared to conventional machine learning models.

There has been a widespread misconception that the role of physicians in healthcare systems is limited to accurate diagnosis and prescription writing. This poor vision is based on the assumption that the patient will fully adhere to the written medical prescription, which rarely happens in reality, because most patients disregard their physicians’ instructions for purposeful reasons like financial hardship or inadvertent causes like forgetfulness. In the contemporary university community, which blends in-person instruction with distance learning, the duties of University faculty members go beyond simple research and teaching to include other responsibilities that would place more burdens and stress on them, which could have a detrimental effect on their lives and cause their medical treatment regimens to fall flat totally. With the development of artificial intelligence techniques and the increasing use of mobile devices, it's easier to develop intelligent apps that cover every part of our everyday routine, including the medical sector, as it's now possible to remotely diagnose, treat and monitor patients’ adherence to prescribed medication plans without the need for direct human involvement. This paper combines artificial intelligence techniques and mobile technology to build a healthier university community by providing an effective smart medication reminder mobile app that supports the principle of medication self-management to improve adherence of medication in-take among patient faculty members at Mansoura University who are undergoing long-term therapy. The evaluation plan of the proposed smart medication reminder mobile app was implemented at two primary levels. The proposal’s acceptability was tested at the initial level by a team comprising both mobile app developers and medical professionals. The proposal’s feasibility was tested on a random sample of patient faculty members from Mansoura University in the second level. The outcomes of the first evaluation level showed that, the services provided by the proposal were highly gained satisfaction of the evaluation team, which means it is suitable for wider use in University environments. While, the outcomes of the second evaluation level revealed that the percentage of taking meds improved among the sample of patient faculty members after using the proposal more than before, which means that it is a useful tool to enhance medication adherence of patient faculty members, especially the elderly with chronic medical disorders.

Prediction of time series is a vital issue related to an extensive array of financial, and social applications, and engineering. The main challenge arises from the intricacy due to the temporal assets of time series and the unavoidable weakening function of analytical systems. Therefore, it is usually problematic to precisely forecast values, particularly in a multi-step ahead situation. Multi-step financial stock price forecast over a lasting perspective is vital for predicting its instability, letting economic organizations charge and evade derivatives, and banks to measure the hazard. Recently, Deep learning systems have been capable to perceive and analyze intricate patterns and connections in the data automatically and haste up the trading procedure. This manuscript designs and develops a Multi-Step Financial Stock Index Forecasting Model Using a Convolutional Neural Network with Gated Recurrent Unit (MFSIFM-CNNGRU) model. The proposed MFSIFM-CNNGRU model relies on enhancing the predicting model for the financial stock index. To accomplish that, the data normalization stage is initially performed by employing z-score normalization to convert input data into a suitable format. Next, the proposed MFSIFM-CNNGRU model designs a hybrid of convolutional neural network and gated recurrent unit (CNN-GRU) technique for the prediction model. Eventually, the hyperparameter selection of the CNN-GRU model can be implemented by the design of the improved whale optimization algorithm (IWOA). The efficiency of the MFSIFM-CNNGRU method has been validated by comprehensive studies using the benchmark dataset. The numerical result shows that the MFSIFM-CNNGRU method has better performance and scalability under various measures over the recent techniques

Deep learning (DL) is recognized as a breakthrough in the educational technology arena, more so in the sense that it can be applied for forecasting student performance and critical issues in academic systems. This systematic review is used to investigate advances in the DL-based system-to-predicting student performance and emphasizes its applicability, methodologies, and limitations. The paper analyses key technologies such as neural networks (NNs) and ensemble models used in educational data mining. The paper also points out limitations in previous studies, for example, data imbalance model interpretability, and issues of scalability. This review highlights the potential of DL to improve educational quality, provide personalized learning experiences, and mitigate learning hazards by synthesizing ideas from different studies. Future directions will comprise hybrid models, improvements in data preprocessing, and merging with real-time educational systems to optimize the performance of the prediction model in several academic environments. For this review, 58 papers were collected from the year 2017-2024 respectively based on DL in education, Risk in education, and student education performance analysis. Subsequently, the aim, technique used, dataset used, performance score attained, significance, and limitations of the existing studies were discussed in this review.

Financial markets are an intricate dynamic system. The difficulty comes from the contact among a market and its applicants, which means, the integrated consequence of the activities of whole applicants decides the market trend, while the market trend disturbs the actions of applicants. These linked interactions make financial markets keep developing. Financial markets are interchange financial instruments like savings certificates, bonds, stocks, and much more. Particularly in stocks, because variations in stock prices are inclined by numerous factors, with economic cycles, financial trends, financial structure, and other macro issues, as well as industry growth, listed businesses’ financial quality. In the last few years, deep learning (DL) and machine learning (ML) techniques have been very effective in predicting financial futures. This study develops a Multi-Criteria Decision Support System for Predicting Financial Futures Using Ensemble of Deep Learning Algorithms with Heuristic Search Mechanisms (MDSSPFF-EDLAHS) model. The main intention of the MDSSPFF-EDLAHS method is to predict future of finances using advanced ensemble models. At first, the data normalization stage applies min-max normalization for transforming input data into a beneficial format. Besides, the ensemble of deep learning models namely variational auto encoder (VAE), bidirectional long short-term memory (Bi-LSTM) technique, and dueling double deep Q-network (DDQN) system have been executed for the prediction of financial futures. At last, the spider wasp optimization (SWO) algorithm adjusts the hyperparameter values of the ensemble models optimally and outcomes in greater prediction performance. The experimental evaluation of the MDSSPFF-EDLAHS is examined on a benchmark dataset. The extensive outcomes highlight the significant solution of the MDSSPFF-EDLAHS approach to the financial future predicting process