With the ever-increasing volume of online information, recommender systems (RS) have been an effective technique to overcome such information overload. Given its widespread implementation in a variety of web applications, the value of RS cannot be emphasized enough. RS has the potential to improve numerous issues identified with over-choice. Deep learning's influence is also widespread, with new evidence of its efficacy in information retrieval and RS research. In RS, the field of deep learning is emerging. This article presents a thorough evaluation of recent research findings on traditional RS and deep learning-based recommender systems (DLRS). This review covers articles published between 2018 and 2022 in four major research databases: Science Direct, IEEE Explore, Springer, and Wiley. First, we provide a complete overview, comparison, and summary of traditional RS. We then systematically analyze DLRS challenges and related solution approaches. Finally, under the discussion of open issues at DLRS, we highlight possible research directions in this area. In this review, we report quantitative data from previous studies to highlight comparisons based on metrics such as F1-score accuracy, recall, accuracy, and error functions.
Selecting the cluster head over the WSN is a challenge. In the existing methodologies, the CHs require high residual energy and also reduce the quality of the system. So, to keep the system more efficient as well as to enhance the system quality a novel fuzzy-based sea lion optimization (FbSLO) mechanism has been developed in this research. Here, the cluster head selection is completely based on a fuzzy algorithm. Moreover, based on sea lion optimization shortest transmission path is measured to transmit the collected node information to the Base station (BS). Subsequently, the performance of the proposed model can be measured in four different parameters as Throughput, Network lifetime, Delay, and Residual energy. While comparing the performance score of the proposed model with the existing models, the developed model attains a better rate of performance. The proposed model attains high throughput is about 75 Mbps and the system has a poor delay is about 55 ms.
Invasive ductal carcinoma (IDC) is the type of breast cancer that occurs most frequently. IDC detection is a laborious process and important task. Furthermore, accurate IDC detection is necessary to provide patients with appropriate medical care. Deep learning technique is far more reliable in these kinds of medical scenarios. This study proposes a deep learning model for identifying IDC in breast cancer histopathology images as a tumor is malignant or benign. Residual Network (ResNet) is a component of our deep learning model. Extract the similar features and train to identify future images; this residual network is given tagged BH image data. As a result, it can recognize the presence of IDC in images of breast tissue lesions. The proposed model produces higher accuracy than the existing models.
Under this work, we propose a novel method for analyzing heart sounds to find cardiovascular illnesses. The phonocardiography (PCG) signal is divided into four pieces by the heart sound segmentation process: S1, systolic interval, S2, and diastolic interval. One may argue that it is an essential stage in the automated analysis of PCG signals. Mechanical activity of the circulatory system is conveyed via heart sounds. This data comprises the subject's precise physiological condition as well as any short-term variations connected to the respiratory cycle. This paper's focus is on an issue that is currently open: how to analyze noises and extract physiological state changes while keeping short-term variability.
Vehicular ad hoc networks (VANETs) are one among the trending research area, and it is utilized in maximum of the intelligent transmission system (ITS) based application. The major challenges which are present during the process of communication between the vehicles are delay due obstacles in ground level and packet loss due to attacks. As so to overcome this drawbacks, hybrid security-based drones-assisted VANETs are introduced in this paper. The proposed approach is based on three transmission models such as (i) vehicle to vehicle (V-V), vehicle to drone (V-D), and drone to drone (D-D). Drones are launched in the VANET network to create the obstacles-free communication. Hybrid security is performed in two stages, at the initial stage—trusted flag-based security is provided to both the V-V and V-D communication models. Secondly, trust evaluation is performed in D-D communication through trust score calculation. To analyze the performance of the hybrid security approach, black hole and Sybil attacks are introduced in the network. The parameters which are considered for the performance analysis are packet delivery ratio, throughput, packet loss, and delay. The values are calculated, and it is compared with the RSAH method in terms of black hole and Sybil attack. Finally, the proposed hybrid security approach achieves (9
