This paper presents systematic literature survey on meta-heuristic algorithms that provide solutions for the multiclass feature comprehension problems in machine learning. Primary behaviors such evolutionary, swarm-intelligence, physics and human-life are the phenomenal and compositional structures in meta-heuristic algorithms. The present review examines the variants of multiclass feature selection, variable classifiers and other application areas. The current article focuses on the certain challenges in meta-heuristic algorithms and identifies gaps useful for the future research studies.
A key component of quantitative brain image interpretation is the precise classification of brain tumors in magnetic resonance (MR) images, which has attracted a lot of scientific interest. Conventional techniques for segmenting MR brain images, including K-means and fuzzy C-Means (FCM) algorithms for clustering, handle every pixel independently and do not integrate spatial information between nearby pixels. Therefore, the noise and degree of homogeneity in brain magnetic resonance imaging poses challenges to the precision of these classification methods. We present a novel way to classification that tackles this problem by utilizing the hybrid shaft clustered technique, that blends Fuzzy Kernel C-Means (FKCM) and adaptive K-means clustering. Furthermore, we determine the area by figuring out how many cells the tumor occupied and how segmented its area is. The results of our simulation show that, in comparison to traditional methods, our suggested method delivers greater segmentation and area estimating reliability.
To track a whole area, an enormous amount of sensor nodes are sprinkled throughout it in a wireless sensor network. The sensor node itself is rather small and has severely limited memory, computing power, and battery life. The wireless sensor network must diminish its energy consumption so as to prolong the system’s lifespan due to the batteries’ limited power supply. An enhanced version of the LEACH routing protocol is offered in this study as a result of a cost-efficient cluster head selection that permits the two variables distance and energy to be major factors. The obsolete LEACH routing methodology is immobile represented for all of the study fields. Clustering-based protocols can adjust energy consumption by giving each node an equal chance to turn into the cluster head. In this study, we deliberate on more modern hierarchical routing algorithms, that rely on LEACH protocol to improve their functionality and prolong wireless sensor network lifetime. Consequently, the most energy-efficient cluster head for data connection is selected depending on the energy and distance characteristics. The simulation’s findings demonstrate that, when compared to the LEACH technique presently in use, the updated approach’s choice of cluster heads is noticeably better. The core cluster head selection procedure, that in various LEACH versions causes an unanticipated breakdown for some cluster heads, can be overcome by the suggested technique. Furthermore, compared to previous iterations of the LEACH protocols, the method offers satisfactory consequences in terms of the network lifetime and energy usage.
Some of the activities encompassed in this category include the practice of business profiling, the implementation of accountability care systems, and the adjustment of capitated scientifically pricing models using patient spending estimates. The present systems in use have significant challenges related to their predictive capabilities and data requirements. These approaches heavily depend on manually constructed models based on point and linear regression techniques. This article presents a multi-view deep learning system that use past claims data to forecast individual-level healthcare expenses. Our multi-view technique is capable of successfully representing several types of data, including patient demographics, clinical codes, drug execution, and facility utilization. Budget prediction study was conducted with data obtained from an authentic paediatric database. The strategy we offer demonstrates superior performance compared to all baseline methods in accurately estimating research expenditures, as supported by empirical evidence from the obtained data. The results of this research enhance the advancement of healthcare by enhancing approaches for both preventative and curative measures.
The current research challenge is to build algorithms that are output-focused and adhere to the trend of quality assurance. Inter- and intra-behavioral changes in the system have an impact on the output quality. This is a result of the class imbalance problem (CIP) that exists in the data sets and is caused by the reading of unknown values. When data sets are in such a state, predictive models that rely on them become severely unbalanced. As selection of samples and candidates for the algorithms is the first step in the CIP on uncertain data, the chosen random samples predict a high degree of class variation, a very small class, or perhaps a class that is much larger than anticipated. The semantic segmentation of images can be accomplished using a dynamic weighting method that is based on the effective sample idea. The results show that this weighting approach can improve the minimal-class segmentation accuracy while guaranteeing that the segmentation performance overall in multi-class segmentation tasks is verified in two separate semantic segmentation tasks.
Since the Industrial Internet of Things (IIoT) is being used widely across many industries, protecting systems used in the IIoT has grown crucial. In order to protect internet of things networks from criminal activity, intrusion detection is essential. In this paper, we offer a structure for deep learning for IIoT intrusion detection that depends on Convolutional neural networks (CNNs). We make use of the reference data set NSL-KDD, which is frequently used to assess intrusion detection systems. The suggested architecture eliminates the requirement for human feature engineering by utilizing CNNs’ built-in capacity to obtain attributes from network traffic data. The dataset is preprocessed, augmented with new data, and divided into training and testing sets. Next, the training set is used to train the CNN model, and the testing set is used to assess it. Our CNN-based intrusion detection system outperforms conventional machine learning techniques in regards to precision, recall, precision, and F1-score, as shown by experimental findings. The suggested architecture offers an effective way to find intrusions in IIoT systems, strengthening their resilience and safety.
The classification of forest cover types is a vital component of environmental management, conservation, and ecological studies. Forests, with their diverse flora and fauna, play a pivotal role in maintaining the planet’s ecological balance. Therefore, precise, and efficient forest type classification based on spectral characteristics is of paramount importance. In this research, an autoencoder-based deep architecture is proposed for forest cover type classification. This research seeks to address the challenges inherent in leveraging spectral data for this purpose. Spectral data is rich with nuanced information that can be challenging to interpret and classify accurately. The Autoencoder-Based Architecture, a deep learning model, demonstrates ability to decipher these intricate spectral features. Autoencoder-based architecture demonstrated a classification accuracy of 98.2
Nations of this century shall improve emergency communication networks complying with the present industry standard. Communication and Technologies recommend a stable emergency communication network mitigating with majority of natural disorders, accidents a stable strengthened emergency communication networks. The paper proposes a broadband-narrowband integrated network for visual dispatch of rescue services with a comprehensive analysis of emergency communication networks, focusing on satellite, ad hoc, wireless and cellular networks. The survey presented examines current emergency rescue networks, future directions for communication networks, and ponders to develop a network resilient to disasters and accidents.
In current research, scholars have focused their efforts on employing machine learning methodologies to analyze multi-media data in order to classify searches made by users. In order to develop a dependable categorization algorithm that functions effectively inside a high-dimensional space, a combination of a robust classification and an extensive feature extractor is employed. In this paper, the integration of three distinct methods is employed to develop a probabilistic belief space policy. The techniques under consideration are assumption space planning, maximum entropy reinforcement learning (ML-RL), and generative adversary modeling (GAM). The offered techniques provide an assessment of different unmodeled adversarial techniques in order to attain robustness. This is motivated by the fact that the simulation include malicious behaviors. The aforementioned framework is employed with the intention of diminishing the agent's capacity for action dependability. The utilization of the reinforcement-based Deep Learning (DL) technique has the potential to be applied in the context of multi-model classification.
Soil is one of the most important resources that we must carefully manage. The fertility of the soil affects crop output by around 60
The cyber security business has greatly benefited from the use of artificial intelligence (AI) methods, which offer the potential for highly self-disciplined models capable of detecting and preventing assaults. The Intrusion Detection System (IDS) is a network security solution that was initially developed to identify and detect instances of vulnerability exploits targeting a specific application or computer system. In order to mitigate cyber attacks, the development of Intrusion Detection Systems (IDS) has been undertaken. Currently, our research employs the support vector machine (SVM), Random Forest, and artificial neural network (ANN) techniques. Based on the findings, the accuracy of Support Vector Machines (SVM) was determined to be 97.800005
Monkeypox, a rare yet emerging viral disease, presents a significant public health challenge due to its potential for human-to-human transmission and clinical resemblance to other skin conditions. Early and accurate detection of Monkeypox is crucial for timely intervention and containment. This research addresses this pressing issue by proposing a modified Inception-v3 deep learning architecture tailored for the detection of Monkeypox from dermatological images. This research is motivated by the importance of early diagnosis in mitigating the impact of Monkeypox outbreaks. The modified Inception-v3 model is customized to capture features specific to Monkeypox skin lesions, enhancing its diagnostic accuracy. Comprehensive experiments were conducted comparing the performance of proposed model against commonly used deep learning architectures, including ResNet, VGG16, and Xception. Experimentation is performed using a real-world dataset. The results demonstrate superiority of proposed model in terms of accuracy, precision, recall, and F1-Score. This research is targeted to contribute to field of medical image processing by addressing the gap in research dedicated to Monkeypox diagnosis.
When it comes to maintaining a growing population and supporting a healthy economy, the agriculture sector is primary. Infectious plant diseases threaten biodiversity and could result in crop losses. In agriculture, early disease detection and identification from leaf photos using machine learning is a tough but crucial topic of study. Since agriculture is a significant economic factor in India (17
Skin cancer is a fatal disease, requiring an automated mechanism for early and accurate diagnosis. In this paper, an approach utilizing the power of Swin Transformers, a deep learning architecture, to detect skin cancer has been proposed. Swin Transformers are recognized for their ability to capture complex patterns and dependencies in high-resolution images. Considering this aspect a swin transformer based model with a transfer learning approach has been proposed to detect malignancy of skin lesions. Transfer learning is used to enhance accuracy and helps for faster convergence. The proposed model utilizes fine-tuning of pre-trained Swin Transformer models on dermatoscopic images and achieves an accuracy of 94
The emergence of deep networks has enabled substantial advancements in the pasture of sensory technology. The proliferation of visual media, coupled with the availability of advanced editing software, has led to a considerable enlarge in the ability to modify digital content. The user’s text is already academic in nature. No further rewriting is necessary. In order to detect and recognize fraudulent actions, we have put forth innovative methodologies. In this research, we discuss two fundamental aspects of employing deep convolutional neural networks in the context of image forgery detection. Firstly, an examination is conducted on different pre-processing techniques in combination with the Convolutional Neural Network (CNN) architecture. Following this, we evaluate the effectiveness of several transfers learning methods, including the utilization of pre-trained ImageNet models through fine-tuning. These techniques are applied to our dataset, CASIA V2.0. The focus of our study entails the examination of preprocessing techniques utilizing a straightforward convolutional neural network framework, while also exploring the profound impact of transfer learning models.
Chili plants are prone to a range of illnesses that can have a substantial influence on their overall development and yield. The timely and efficient implementation of management strategies relies heavily on the early detection and correct identification of these disorders. In recent years, there has been a growing body of research that demonstrates the efficacy of deep learning approaches, specifically convolutional neural networks (CNNs), in the field of plant disease identification. The present study introduces a ResNet-CNN classifier as a means of detecting illnesses in chili plants. The ResNet architecture is a convolutional neural network (CNN) model that has exhibited outstanding results in challenges related to the classification of images. The ResNet-CNN classifier effectively utilizes its deep layers and bypasses connections to learn and retrieve high-level features from photos of chili plants. The classifier under consideration has been trained using an extensive dataset consisting of labeled photos that cover a range of diseases affecting chili plants. The photographs have undergone pre-processing techniques in order to improve their quality and standardize their characteristics. The ResNet-CNN algorithm is subsequently developed based on the labeled dataset, adopting various strategies including data augmentation to enhance generalization and alleviate the issue of over fitting. In order to assess the efficacy of the classification algorithm, a distinct test dataset consisting of photos of chili plants afflicted by recognized diseases is employed. The ResNet-CNN classifier exhibits a notable level of efficiency and effectively proves its capability to precisely classify several diseases that impact chili plants. The model that has been trained can serve as a valuable tool for automatic detection and diagnosis of diseases, hence enabling prompt intervention and mitigating losses in crop output.
A malignant brain tumor is a tumor that has spread across the brain and is threatening human health. Correctly dividing tumors into subtypes and classes is essential for later prognosis and therapy planning. Identifying a brain tumor can be a tedious and error-prone process, hence radiologists need to use automation whenever possible. This paper presents conditional deep learning for structural multimodal MRIs of the brain to perform tumor categorization using a residual network, survival rate forecasting, and dissection, to name a few. To begin, we recommend a segmentation method that separates non-overlapping regions using a combination of conditional random fields and convolutional neural networks. Using these patches, finding the tumor takes hardly no time at all. Errors multiply if their scopes cross. In the paper’s second section, the authors provide a method of feature mapping using a residual network and XG-Boost for training models. The following part focuses mostly on these two topics since they are related to the reduction of information loss and the improvement of tumor data quality, respectively, through the use of residual features and nonlinear space mapping. The XG-Boost-learned mapping of features enhances structural-based learning and boosts accuracy across classes. A cancer dataset and a non-cancer dataset, as well as a meningioma, glioma, and pituitary dataset, are used in the experiment. Both areas saw dramatic performance boosts compared to alternative methods. The primary focus of this study is on enhancing segmentation and its implications for measures of classification efficiency. The use of a residual network and a conditional random field helps to improve the quality. The outcome is a 3.4
Clustering techniques assume a major role in mitigating security concerns within cloud computing settings. The objective of this study is to assess and contrast different clustering algorithms in terms of their efficacy in augmenting cloud security. The evaluation encompasses various factors, including the accuracy of algorithms, the time taken for performance, and their efficacy in identifying and addressing security problems. The results indicate that the Make Density Based Clusterer and Simple K Means algorithms exhibit superior accuracy rates in comparison to the other algorithms that were assessed. These algorithms demonstrate a harmonious equilibrium among precision and speed of execution, rendering them excellent selections for augmenting the security of cloud systems. Furthermore, the Farthest First algorithm demonstrates the lowest execution time compared to the other examined algorithms, thus emphasizing its effectiveness in rapidly clustering data. In summary, the empirical findings suggest a notable disparity among the clustering algorithms in terms of their efficacy in identifying and mitigating attacks and vulnerabilities in Cloud-based applications. The observed discrepancies in reliability measurements underscore the divergent efficacy of the algorithms. The Make Density Dependent Cluster and Simple K Means algorithms provide superior reliability, indicating their efficacy in bolstering security inside the cloud environment.
Currently, breast cancer is a significant contributor to mortality among women, ranking second only to lung cancer in terms of cancer-related deaths. Breast cancer arises as a result of the fast proliferation of cells within the breast tissue. Breast cancer can manifest in any region of the breast and can be prevented with early initiation of treatment. Breast cancer is a malignant neoplasm characterized by the uncontrolled proliferation of cells originating from breast tissue. The management of breast cancer is contingent upon the specific subtype of cancer and its corresponding stage, ranging from stage 0 to stage IV. Treatment modalities commonly employed encompass surgical intervention, radiation treatment, and chemotherapy. The objective of this study is to employ various data mining algorithms and technologies to diagnose breast cancer. The identification and analysis of disease patterns in earlier cases have led to the discovery of a novel facet in the field of medical advancement. The dataset used in this study was obtained from the UCI machine learning repository. The research focused on evaluating the classification performance of various algorithms, including Binary Categorization, Pseudo random Forest, Regression Analysis, Multilayer Perceptron, and K-NN, for predicting the occurrence of Breast Cancer Disease. The analysis of the gathered data encompassed an examination of its dependability, true positive rates, F1-score, and Kappa statistics.
In the realm of Vehicular Ad-Hoc Networks (VANETs), the confluence of an open-access data transmission environment and the imperative to ensure privacy and security poses multifaceted challenges. This study delves into the intricate landscape of VANETs, characterized by data transmission vulnerabilities, to address the pressing concerns of user information protection and network integrity. The core emphasis of this investigation lies in fortifying VANET security through a meticulous approach that amalgamates an attack tree-based model with a robust trust-based evaluation system. In striving to augment privacy and reliability within VANET systems, the study undertakes a comprehensive exploration of security requisites encompassing availability, integrity, and confidentiality. Central to the proposed framework is the strategic deployment of an attack tree model, meticulously constructed to encapsulate potential attack strategies. This model is further synergized with a dynamic integration of trust evaluation and clustering mechanisms, culminating in an agile and effective security architecture. Specifically, a VANET system is instantiated with an initial set of 20 nodes, thereafter subjected to varying node densities post simulated attack scenarios. The orchestrated interplay among Roadside Units (RSUs), nodes, and routing protocols affords a visual elucidation of data packet transmission dynamics, concurrently facilitating the prompt identification of potentially malicious nodes. In conclusion, this study contributes a holistic approach to VANET security enhancement, optimizing bandwidth allocation without compromising node integrity. The synthesis of attack tree modeling and trust-based evaluation engenders an overarching security paradigm for VANETs, fostering privacy, dependability, and overall network resilience.
