Advancements in Smart Computing and Information Security

This study focuses on the application of classification rule mining techniques to analyse biological and healthcare data, specifically using a tuberculosis dataset. Naive Bayes, Logistic Regression, Decision Tree, Random Forest classifier, K-Nearest Neighbour, and Support Vector Machine were among the classification techniques examined in the investigation. Support Vector Machine (SVM), Random Forest, and Decision Tree algorithms show the highest degree of accuracy.

Multimodal Sentiment Analysis (MSA) is a burgeoning field in natural language processing (NLP), also known as opinion mining. It determines sentiment(positive, negative, neutral), subjective opinion, emotional tone, sometimes even more fine-grained emotion like joy, anger, sadness, and others. The evolution of sentiment analysis from its early days of text only analysis to the incorporation of multimodal data has significantly enhanced the accuracy and depth of sentiment understanding. MSA is poised to play a pivotal role in extracting valuable insights from the vast amount of multimodal data generated in today’s digital age. Various fusion methods have been developed to combine information from different modalities effectively. Additionally, the field has seen significant contributions from lexical-based, machine learning-based, and deep learning-based approaches. Deep learning, in particular, has revolutionized MSA by enabling the creation of complex models that can effectively analyze sentiment from diverse data sources. This survey provides an overview of the critical developments in MSA, highlighting the evolution of methods. It also presents a comparative analysis of state-of-the-art models and their performance on benchmark datasets and future potential, helping researchers and practitioners choose the most suitable approach for their specific tasks. The surveyed models SKEAFN, TEDT, UniMSE, MMML and others have exhibited impressive performance across various datasets.

Rice is a major crop that has a major impact on the Indian economy. Indian farmers face many financial problems when rice cultivation suffers from diseases that direct to decline of the mixed economy. The most important economic and scientific challenge in agriculture is the categorization and identification of rice disorders. Detection and monitoring of theses disease is the critical issue. If these diseases are identified at the first stage appropriate action could be taken in order to restrain the economic loss of the farmers. To overcome this, four different rice diseases are acquired from the online datasets. All the images are preprocessed, and the extracted features are given as the input to Random Forest. In India, there are four primary rice plant diseases: leaf blast, bacteria blight, sheath blight, and brown spot.

In the whole world, Cardio Vascular Diseases (CVDs) are the main reason of death. The outcomes of patients are significantly improved by early detection and precise prediction of CVDs. We offer an in-depth process for feature extraction and classification for CVD risk identification in this paper. By combining the strength of Support Vector Machines (SVM) classification with Linear Discriminant Analysis (LDA), Principal Component Analysis (PCA) and PCA with ReliefF feature retrieval methods, this study presents an investigation into feature extraction approaches for CVD classification. On a variety of CVD datasets, tests were run to see how well the PCA-ReliefF feature extraction strategy performed when combined with SVM. In this study, we emphasize the significance of not only accuracy but also recall as a key metric, shedding light on the model’s ability to correctly identify individuals with cardiovascular illnesses. The PCA + ReliefF + SVM model outperforms other algorithms with a consistently higher accuracy ranges from 91.4% to 93.4%, a recall between 82% to 84% and precision ranges from 82% to 84%. The language used for execution is Python.

In this research process, a medical decision model is developed for disease prediction based on DL (Deep Learning) models. The major benefits of computer-based algorithms are exact results, adaptability, transparency, and better decision-making. The proposed work three major steps are preprocessing, feature selection and classification. Firstly preprocessing, data analysis pre-processing is the major step in identifying exact methods. Most of the clinical data consists of missing information and inconsistent data. WB-SMOTE (Weighted Borderline Synthetic minority oversampling technique) concept is applied to asses and solves the unbalanced. Secondly feature selection, selections of features are the process of choosing a subgroup of the most associated attributes in the concerned dataset to indicate the final identifier. Wrapper-based approaches are applied to extract the features from the given dataset. Finally classification, accurate prediction of diabetic disease based selected number of features. Classification approaches are Decision Tree (DT), Random Forest (RF) and Enhanced Convolution Neural Network Layer (ECNN). The output comparison among the DRL-OCNN model and some other ML Models is offered. While analyzing diabetes data, it is identified that DRL-OCNN models produce better results with 95.75% of accuracy rate. The received results demonstrate that this suggested DRL-OCNN model produces better performances with a precision of 0.93 and recall of 0.91. This enhancement can decrease time, labor services, effort, and decision exactness. The planned system was assessed on PID (Pima Indians Diabetes) and illustrates an excellent performance in forecasting diabetes illness. The tool used for execution is python.

Kidney diseases are commonly viewed among people. Medical analysis of Chronic Kidney Disease (CKD) is performed with a blood test and urine test. In recent times, data mining and analysis concepts are implied for predicting CKD through the application of patient details and recorded data. At this moment, predictive analysis modeling such as Support Vector Machine (SVM), Multilayer Perceptron (MLP), Linear Regression (LR) and proposed Optimized Multi-Layer Perceptron (PKD-OMLP) is executed for predicting CKD. Pre-processing is employed for reducing the level of misplaced data and impure data. During the processing stage, the identifiers are spotted which aid in the model forecasting. The selected three types of predictive algorithms are assessed and appraised relying on their prediction accuracy, precision values, and recall. The research study provides a decision-making tool that supports the forecasting of kidney diseases. The main goal of the study is to recognize CKD diseases at an earlier stage with the assistance of Machine Learning (ML) models like Linear Regression (LR), Support Vector Machine (SVM), and Multi-Layer Perceptron (MLP). In this study, models are designed with the use of Python programming with Python 3.7.0 and their performance is contrasted concerning the recall, accuracy rate, and precision. Among the preceding four models PKD-OMLP gives the best outcome as per its performance level producing accuracy of about 94.75%, precision of about 0.93 and recall of about 0.92 respectively on testing CKD dataset from Kaggle.

Machine learning (ML) based heart disease prediction has emerged as a crucial and fruitful field of study and application. They are used to analyze medical data, identify cardiac disorders, and make precise predictions about their presence or absence. Utilizing the effectiveness of LR (Logistic Regression) in combination with sophisticated preprocessing methods has emerged as an important strategy in the classification of heart-related diseases. By addressing data variability and differences in feature sizes, the use of decimal scaling and min-max normalization, in particular, improves the interpretability and flexibility of the model. By detecting complex underlying patterns and nonlinear interactions within the data, the use of Isomap (Isomeric Feature Mapping) normalization enhances the LRs (Logistic Regression) discriminative powers. This Scope of research has shown the possibility to produce higher classification results through a thorough review procedure that includes accuracy, precision, and recall criteria. Isomeric Feature Mapping along with LR gives best result with accuracy of 91%, precision of 0.89 and recall of 0.87 respectively. This proposed system is compared with the existing methods like Min-Max Normalization+ LR, and Decimal Scaling Normalization+ LR. The tool used for execution is python.

Automation in the agriculture field is a priority when compared with other fields, as the latest growth of Agriculture and Farming is dependent on technologies for production. The next major important requirement in Plant diseases is early prediction and necessary related recommendations. In this research, the proposed method is implemented with the plant leaf dataset, which consists of diseased and healthy data of various plant leaves. The prediction and classification of the diseased plant leaves is achieved by deploying the deep neural network models ResNet50, AlexNet and Proposed model ProliferateNet. Finally, the experimental output values of these models show the significance of the Neural Network models in the detection of plant disease, as well as the efficiency of neural networks. During training a Neural Network model, data augmentation can solve a number of issues, including limited or imbalanced data, overfitting, variance, and complexity. The dataset is augmented using image-based data augmentation techniques before being applied to deep neural network. The accuracy of the various models is evaluated, and ProliferateNet attained an average training accuracy of 93% and testing accuracy of 99%.

Cancer treatment is conceivable on the off chance that can ready to identify it at a beginning phase. For the most part, Side effects of disease are found in human body in last stage, however with assistance of trend setting innovation where PC supported frameworks are utilized; we can identify it in a beginning phase. Right now, various AI strategies are utilized for such computerized location frameworks to distinguish cellular breakdown in the lungs in beginning phases. For such computerized identification, we utilized CNN and CT images. Using DL methods, this study enhances a novel method for Computer tomography and Magnetic resonance image-based lung tumour detection feature selection and extraction. The CT and MRI lung images that were used as input were processed for noise removal and normalization. Following that, a gradient support vector discriminant neural network and kernel convolutional component analysis are used to features selection with feature extraction from the processed images. The experimental analysis is carried out based on parameters Random accuracy, F-1 Score, mean average Precision (mAP), dice coefficient, kappa Co-efficient for various MRI and CT image dataset. Performed algorithm had Random result of rightness 95%, 75% of F-1 score, mAP of 81%, dice coefficient of 68%, kappa Co-efficient of 55% for MRI image and Random accuracy of 96%, F-1 Score of 66%, mean average Precision (mAP) of 55%, dice coefficient of 68%, kappa Co-efficient of 63% for CT image.

Numerous nations have enacted total lockdowns in an effort to contain the Covid-19 pandemic, which is spreading quickly throughout the globe and claiming millions of people every day. As people tended to vent their emotions through social media during this time of lockdown, these channels were crucial in helping to distribute information about the pandemic around the globe. We created an experimental methodology to examine Twitter users’ reactions while taking into consideration the terms that are frequently used to refer to the epidemic, either directly or indirectly. In order to carry out the text classification, the TF-IDF method is upgraded (TF-IDCRF) in this study. The dataset involved with 44,995 tweets from all over the world and the DL approach is utilized for improving classification accuracy by addressing the issue in inadequate classification of feature category. Finally, the suggested approach is compared to two DL methods with TF-IDF algorithms with Long Short Term Memory (LSTM) and Gated Recurrent Unit (GRU) and the better prediction of tweet category is determined in which GRU performs high accuracy as 92.4% than LSTM technique.

Cancer Blood Disorder has an impact on the development and operation of our blood cells. Blood disorders can affect platelets, blood plasma, white and red blood cells, or any one of the four main components of blood. Proposed work goal is to identify cancer blood condition. In this research, Images of cancer and blood disorder are preprocessed utilizing enhancement and filtration methods. In research suggested a 2D Hybrid Wavelet Frequency Domain Bilateral Filter (2D HWFDBF) for noise removal. To increase the clarity of an image, image enhancement is used. Apply, proposed a 2D Edge Preservation Histogram Improvement (2D EPHI) technique for image enhancement. Real time data set was collected for image preprocessing. The proposed filtering technique is very effective and produced the best result when compared to the other filtering techniques such as 2D Hybrid Median Filter, 2D Adaptive Log Color Filter and 2D Frequency Domain Filter. Proposed image enhancement technique carried out the best outcome when compared to the other techniques such as Contrast Limited Adaptive Histogram Equalization, Image Coherence Improvement and 2D Adaptive Mean Adjustment. MATLAB software can be used to implement the proposed system. To evaluate proposed system by using RMSE (Root Mean Square Error) and PSNR (Peak Signal to Noise Ratio). These outcomes are compared to the existing methodologies. Finally, results of filtering and enhancement techniques shows the better outcome than compared to the existing approaches.

The amount of oxygen present in water, known as Dissolved Oxygen (DO), is impacted by a range of physical, chemical, and biological factors. This measurement is pivotal for assessing the condition of water, as it directly reflects the ability of aquatic ecosystems to sustain marine organisms. Evaluating water quality frequently involves the use of Chemical Oxygen Demand (COD). In the context of facilities treating wastewater, a combination of biological, physical, and chemical techniques is employed to manage industrial waste and remove contaminants before they are discharged into water bodies. Discharging untreated industrial waste into natural water sources is a major cause of water contamination. Standards mandate that the concentration of DO in waste should exceed 3 mg per liter. However, industries aim to maintain low DO levels to minimize the risk of pipe corrosion. Due to the time-consuming nature of manual COD measurement, industries often neglect to check DO levels before disposing of waste. The proposed study seeks to forecast the COD of treated waste from a wastewater treatment plant by utilizing crucial data gathered by sensors from the initial waste. This approach ensures that industries undertake suitable waste treatment before disposal, safeguarding marine life and enhancing the quality of water accessible for daily use.

India’s economy is built upon agriculture, which offers the common of the country’s inhabitants with an existing and accounts for 40% of the nation’s overall GDP. Agriculture is a major component of an argo economy like India’s. The Indian economy benefits from the agricultural sector as well as the industrial sector and foreign import and export trade. Even while the agricultural sector in India currently employs the most people nationwide, its contribution to the economy is shrinking. Cotton is one of the most important commercial crops grown in India; it accounts for about 25% of all cotton produced globally. It is a major source of income for about 40–50 million workers in sectors like trading and cotton processing, as well as 6 million cotton growers. The objective of this article is to provide a summary of the machine learning techniques used to identify and predict a variety of diseases in cotton crops using machine learning and artificial neural networks. An article has thoroughly examined numerous machine learning algorithms and their uses in the field of agricultural disease for this goal. The study also shows how machine learning methods are used in the subject of agricultural disease identification in a comparative and comprehensively tabular manner. In the field of cotton crop disease identification, the article also covered the potential application of machine learning algorithms in the future.

Nowadays many students get enrolled in schools and colleges for their academic career. Early identification of students at danger level, alongside precautionary measures, can completely work on their richness. Recently, ML methods have been widely utilized in the education domain to forecast the performance of students. Predicting higher education rates using machine learning can be approached in several ways, based on the existing data and the definite factors being considered. In this paper, pre-processing, selecting features, reformulating the problem, learning the model, predicting performance, and analyzing results has been used as major steps. SVM, RF, and CNN approaches are applied to prognosis the performance of the learners. The suggested model is designed using Python software and the accuracy of the models is compared. Among the three models, CNN can produce a better result by giving accuracy of about 90.75% and Precision and Recall of about 0.90 and 0.88. Predicting higher education rates using machine learning can provide valuable insights into future trends and help stakeholders.

The most hazardous disease the globe is now dealing with is cancerous. It is challenging to find cancerous nodules inside the lungs, although many techniques have been used to do so. Lung cancer segmentation is a process of identifying and isolating lung cancer tissues from medicinal picture like CT or MRI scan images. This process is essential for accurate diagnosis and management planning of lung cancer. Computing techniques can be used to automate and increase the accuracy of lung cancer dissection. Deep Learning (DL) is a popular technique used in medical image analysis. It has become increasingly important in lung cancer segmentation is the main research work nowadays. This study applied three DL approaches like U-Net, V-Net and the Mask R-CNN for lung cancer separation. Among the three techniques, the U-Net model provides better outcomes based on their evaluation metrics like Accuracy, Sensitivity and Specificity. From the results obtained the proposed U Net gives accuracy of about 97% to 98.4%, Sensitivity of about 88.3% to 91% and Specificity of about 93.2% to 94.6% respectively. The tool used for execution is Matlab.

Skin cancer is a prevalent and potentially fatal disease. Early detection is important for successful treatment. Traditional methods face challenges in identifying skin cancer regions. CSDM-Deep-CNN is a novel approach for efficient skin disease detection with minimal execution time. CSDM-Deep-CNN leverages deep convolutional neural networks with batch normalization. The objective of this study is to address the complexities in dermatology and the increasing impact of skin disorders on individuals’ psychological and social well-being. The proposed CSDM-Deep-CNN approach offers a promising solution by leveraging machine learning and deep learning technologies. The CSDM design and implementation involve pre-processing steps, image resizing, and the use of convolutional neural networks for disease prediction. The optimization process includes batch normalization to prevent overfitting, enhancing the training efficiency of the deep convolutional layer. The study reports promising results, including an accuracy rate of 84%, a training time of 1.59 s, and a total execution time of 4.23 s.

Accurate diagnosis of skin lesions is crucial for early detection and effective treatment of dermatological conditions. However, blurry artifacts present in dermatological images can significantly hinder diagnostic accuracy. Existing research primarily focuses on either shape analysis or deep learning techniques individually, with limited consideration of hybrid approaches that can leverage the complementary strengths of both methodologies. To address this research gap, we propose a novel hybrid blur detection method for enhancing skin lesion diagnosis. Our approach integrates shape analysis techniques with deep learning methodologies to improve the accuracy of dermatological image analysis. Shape analysis algorithms capture intricate shape features of skin lesions, which are then utilized by a deep learning model trained on a diverse dataset of dermatological images. Experimental evaluations demonstrate the effectiveness of our hybrid approach in accurately identifying and localizing blur regions within skin lesion images. By mitigating the impact of blurry artifacts, our method enhances image quality and facilitates accurate analysis, enabling early detection and intervention for improved patient outcomes. This research contributes to the advancement of skin lesion diagnosis by providing a robust tool for clinicians and dermatologists. The proposed hybrid blur detection method has the potential to significantly improve the precision and reliability of dermatological image analysis, leading to more accurate diagnoses and timely treatment decisions.

A common severe phase of diabetes mellitus known as diabetic retinopathy (DR) results in anomalies on the retina that affect eyesight. The likelihood of visual deterioration will be greatly lowered by early identification and treatment with DR. Because of the complexity of imaging environments, fundus images are usually hampered by noise and poor contrast problems. This study proposes an algorithm for enhancing image quality by lowering noise and enhancing contrast. For the purpose of de-noising and enhancing a color fundus image, the incorporation of proposed Edge Preserving filters and Swarm Based Enhancement Optimization method is implemented. A common public dataset called DIARETDB0 is used to assess the experimental findings. The Mean Square Error (MSE), Peak Signal to Noise Ratio (PSNR) and Structural Similarity Index (SSIM) which have been measured as 0.000121, 42.37 and 0.999 respectively, are three performance parameters been used. In comparison to other filtering techniques, the suggested algorithm demonstrated improvement in optimizing the quality of images. The tool used for execution is MATLAB.

Cyber-attacks are getting more and more complicated, using intricate patterns that are challenging to find using conventional techniques. IDS (Intrusion Detection System) are essential for defending computer networks from online risks. This article undertakes a thorough review of three preprocessing methods used with a Convolutional Neural Network (CNN) for intrusion detection along with SMOTE, Z-score, and IQR (Inter Quartile Range) which will be used for feature extraction. The study carefully evaluates the evaluation parameters such as accuracy, precision, and recall, to ascertain the most efficient preprocessing approach. When dealing with sequential data in intrusion detection systems, utilizing a CNN to classify intrusion is a potent technique. The accuracy and dependability of an intrusion detection model can be improved by combining CNN with preprocessing methods like IQR. By handling outliers using the IQR approach, the CNN model is trained on a more accurate and reliable dataset. From the results obtained proposed IQR+CNN produces Accuracy of 90.3%, Precision of 0.90, Recall of 0.87 and F Measure of 0.9. The tool used is Jupyter Notebook and language used is python.

Heart disease and other cardiovascular disorders continue to be the most prevalent cause of death. ML (Machine Learning) algorithms in particular have shown promise in forecasting for early identification and prevention. Using innovative preprocessing methods like Z-score normalization, IQR outlier handling, and Synthetic Minority Over-sampling Technique (SMOTE) for class imbalance, the present research investigates the use of ANN (Artificial Neural Networks) in the early detection of cardiovascular disease. When compared with various preprocessing techniques, SMOTE and ANN regularly exceed them in terms of precision, sensitivity, and specificity, according to the results of the study. The balanced illustration of both positive and negative cases in the synthesized dataset gives the NN (Neural Network) a more thorough learning experience. Since there are fewer false negatives (greater sensitivity) and false positives (more specificity) due to the ANN model’s increased accuracy for forecasting heart disease, there are fewer false positives as well. From the results obtained proposed SMOTE+ANN produces Accuracy of 91%, Specificity of 0.86 and Sensitivity of 0.91. The tool used is Jupyter Notebook and language used is python.

Wind speed, solar radiation, and weather conditions are famous and extensively used RE sources in the global. As a result of their high carbon content and the processes used to produce them, fossil fuels like coal, natural gas, and petroleum cannot be replenished and are therefore not considered renewable energy sources. Demand forecasting heavily depends on irregular renewable sources, whose production is weather-dependent. It was carried out using Machine Learning (ML) techniques. However higher computational complexity and incapability are major important issues of ML methods. This study proposes a new algorithm to use weather forecasts and data on consumption and generation to generate energy demand. Utilizing a model that extends beyond the upcoming day-ahead auction, hourly electricity price forecasting is done. Initially data normalization is used to pre-process the dataset. Then, Entropy Binary Dragonfly Algorithm (EBDA) was introduced to select the most important features at the same time as enhancing the prediction accuracy. Finally, the Optimized Stacking Hermite Polynomial Neural Network Ensemble (OSHPNNE) model is introduced for RED forecasting. HPNN parameters are optimized using EBDA to increase prediction accuracy and enhance classification capacity. Kaggle is used to collect hourly energy demand generation and weather datasets, which have been employed in experiments. Determining the electrical components by extrapolating them based on the influence of weather forecasts on their time, location, and climate. Metrics like Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Pearson Correlation Coefficient (r), and Nash Sutcliffe Efficiency (NSE) have been used to assess the results of forecasting approaches.

Preprocessing is very important to predict Intrusion Detection System (IDS) with respect to any parameters. It entails prepping and converting raw data into a format compatible with Machine Learning (ML) algorithms. ML approaches are used to categorize network activity as either legitimate or malicious to create IDS. For binary classification problems like intrusion detection, one such approach is LR (Logistic Regression). The data must be preprocessed for modeling to be effective. In the present investigation, Min-Max Normalization, SMOTE for controlling class imbalance, and Z-score Normalization were used in conjunction with PCA feature extraction and LR (Logistic Regression) for classification of intrusions. It is possible to considerably increase the accuracy, f1-score, precision, and recall of the IDS by combining the preprocessing method Z-score Normalization for normalization with PCA feature extraction. From the results obtained proposed Z Score + LR produces Accuracy of 88.3%, Precision of 0.86, Recall of 0.84 and F Measure of 0.8. The tool used is Jupyter Notebook and language used is python.

Social Network plays a vital role in exchanging information in this smart world. In such situations, the usage of words is an important one to restrict the abusive information and text in the comment sections. This word usage was restricted by using text mining algorithms and classification techniques. But the word usage was reduced after the emoticons usage. Using emoticons also, the users can convey the harsh comments. As they combine both text and emoticons in their message, the identification of emoticons is important. Based on this, in this work, the identification of emoticons is performed using the Deep learning algorithm called Deep Neural network. Here, the emoticons from different groups were used as the input dataset. This emoticon was processed by using the proposed user defined convolutional neural network layer for emoticon classification. This emoticon classification performance will be analyzed to evaluation metrics. To enhance its accuracy further, the hyper parameters of proposed attention based DNN like learning rate and batch size will be tuned using Particle swarm algorithm. Then, its performance will be evaluated using evaluation metrics for identifying the best deep learning approach for emoticon classification. The whole process will be realized using MATLAB R2022a software.

Due to its non-intrusive aspect and distinctive biometric features, palm print identification technology has attracted a lot of attention recently and is now a crucial part of contemporary security mechanisms. This study explores how to improve palm print image preprocessing methods for security systems using an entirely novel approach called Receiver Operating Characteristic (ROC) assessment. Furthermore, it investigates the extraction of attributes using three well-known techniques: Scale-Invariant Feature Transform (SIFT), Local Binary Patterns (LBP), and Speeded-Up Robust Features (SURF). This study, which highlights its improved performance in terms of Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Silhouette Score, is important for looking at the collaborative influence of ROC analysis in combination with SURF. From the result obtained we can prove that SURF produces MSE of 0.00248, RMSE of 0.05850, Silhouette Score of 0.6, SSIM of 0.998 and PSNR of 42.35respectively which is better than other algorithms. The tool used for execution is Jupyter Notebook and the language used is python.

Worldwide, lung diseases are a common occurrence. It consists of pneumonia, asthma, TB, fibrosis, Chronic Obstructive Pulmonary Disease (COPD), and others. However the early detection of this disease is crucial. Thus several Machine Learning (ML) and image processing methods have been introduced for disease detection from images. Deep learning (DL) is an effective ML approach which integrates the procedure of supervised training by feature distribution and unsupervised training to shorten optimization. Lung disease diagnosis from Chest X-Ray (CXR) images has been extensively studied using the U-NET architecture. Differential Evaluation Multi-Scale U-NET (DEMSU-NET) Architecture, multi-scale feature maps are extracted from every convolutional of the U-NET encoder. Noisy or insufficient annotations may decrease the accuracy of U-NET model; it may be solved by auxiliary confidence maps. It takes place less emphasis on the limits of the provided target detection of lung disease. Differential Evaluation (DE) is implemented to adjust the background and foreground weights based on the population X-ray image. National Institutes of Health (NIH) chest X-ray images are gathered from the Kaggle repository to experiment the detection methods. Results of the proposed system and current methods are assessed using measures such as precision, recall, Fβ-score, and accuracy.

Cyber Hacking Detection System (CHDS) plays a major important role to identify any type of incidents that occur in the system. For instance, a successful CHDS could identify when an invader has compromised a system with the help of the system vulnerability. In addition, many CHDS are capable of monitoring reconnaissance activities, which indicate whether the attack is impending or it is for a particular system or the characteristics of a system that carries specific interests to intruders. The major aim of the work is to design a new CHDS. In this paper, pre-processing SMOTE algorithm and Linear Discriminant Analysis (LDA) by feature selection has been introduced for CHDS. SMOTE preprocessing in Cyber Hacking Detection System (CHDS) can result in a representative and well-balanced training dataset. The LDA method determines a projection vector that decreases the within-class scatter matrix in the feature space while increasing the between-class scatter matrix. For classification, X Gradient Boosting, K Nearest Neighbor (KNN) and Sliding Window based MultiLayer Perceptron (MLP) is used for CHDS. MLP classifier is a set of input-based values to their corresponding outputs. From the results obtained, the proposed Sliding Window based MLP produces Accuracy of 90.70%, Precision of 0.89, Recall of 0.87. The tool used is Jupyter Notebook and the language used is python.

Retinopathy of Preterm (ROP) is becoming more common in babies as the number of preterm individuals grows dramatically around the world. ROP can be effectively treated, but it requires constant screening and early diagnosis. Implementing a computer-aided approach based on image processing is among the simplest ways to diagnose ROP. Deep learning approaches have shown to be quite effective in medical image analysis in this regard. For Noise removal Laplacian of Gaussian (LoG) filter is used. In comparison to Random Forest (RF), Artificial Neural Network (ANN) and Convolutional Neural Network (CNN), the method proposed in this research aims to detect the ROP by using YOLO algorithm to accurately detect and classify retinal fundus images according to its severity. Dataset is collected from Kaggle and the Python package. The experimental studies show that the suggested work is stable, trustworthy, and yields promising ROP detection results with accuracy of 94.63%, sensitivity of 0.94, specificity of 0.80 and F-Measure of 0.8 respectively. The tool used for execution is python.

In the era of information overload, social media platforms like Twitter have become invaluable sources of real-time public sentiment. Sentiment analysis, the process of gauging the emotional tone of text data, plays a pivotal role in extracting insights from these vast repositories of user-generated content. This paper presents a comprehensive exploration of sentiment analysis on Twitter tweets using Hugging Face, a leading natural language processing (NLP) library. This study harnesses the capabilities of Hugging Face’s models, particularly transformers, to perform sentiment analysis on Twitter data. It delves into the methodology of data collection, preprocessing, and model selection, showcasing the versatility of Hugging Face’s transformer models. The practical applications of this research are far-reaching. By analyzing Twitter sentiments, can uncover valuable insights for businesses, policymakers, and researchers. Sentiment analysis on Twitter can help companies gauge the reception of their products or services, enabling data-driven decision-making. Policymakers can utilize sentiment analysis to gauge public opinion on critical issues, aiding in the formulation of effective policies. “HugSent” represents a state-of-the-art sentiment analysis algorithm, leveraging Hugging Face and NLP techniques. This cutting-edge method has demonstrated an exceptional level of accuracy and reliability with perfect precision, recall, F1-score, and support values of 1.00 for sentiment categories, 1 and 0. These refinements aim to enhance its versatility and practicality, catering to industry-specific needs, and making it a more adaptable and nuanced tool for sentiment analysis in diverse contexts.

Cancer is currently the most dangerous sickness the world has to cope with. Finding malignant nodules inside the lungs is difficult, despite the fact that numerous methods have been employed. The process of recognizing and separating lung cancer tissues from medical pictures such as CT or MRI scans is known as lung cancer segmentation. This procedure is necessary for a precise lung cancer diagnosis and treatment planning. Lung cancer dissection can be made more accurate and automated with the help of computing technology. Cancer in lung is one of the primary reasons of demise universal. Timely recognition and accurate diagnosis are critical for improving patient outcomes. In this research study, the authors examine the use of three different DL (Deep Learning) classifiers, namely CNN (Convolutional Neural Networks), RNN (Recurrent Neural Networks), and SAE (Stacked Autoencoders) for the categorization of lung malignancy from CT (Computed Tomography) images. The performance of these models is compared in form of accuracy, sensitivity and specificity. The investigational outcomes show that the CNN model outperformed the other models with an accuracy value of 92.63-%, the sensitivity rate of 0.91 and specificity value of 0.89. The tool used for execution is MATLAB.

The term cyber-attack or intrusion is expanded as an unauthorized process that includes one or more of the above three components of the network system. The intrusion detection (ID) process helps the administrator of the system to build up security mechanisms that recognize the legitimate or illegitimate of the system. The illegitimate user of the network system is named an intruder which can be a person within the organization or outside the organization. IDS are constructed with the concept of observing the unauthorized behavior of the user concerning the authorized behavior activities. The deviation noted based on the comparison is considered an intrusion. Many novel techniques are developed through research to observe and identify the current activities. In this research three algorithms namely R-SVM, Adaptive Boosting and Multi-Layer perceptron have been used. From the results obtained Multi-Layer perceptron produces Accuracy of 92.3%, Precision of 0.89, Recall of 0.87 and F Measure of 0.8. The tool used is Jupyter Notebook and language used is python.

Filtering strategies are frequently used in instantaneous video processes, particularly for applications such as identifying items for traffic recordings, to increase the standard of the footage frames and the precision of object recognition processes. Several varieties of filters can be employed for this, including the Kalman filter, mean filter, and Wiener filter. Key photographic metrics involving PSNR (Peak Signal-to-Noise Ratio), SSIM (Structural Similarity Index), and MSE (Mean Squared Error) are used to evaluate how well these filters work. According to findings from experiments, the Kalman filter operates better based on PSNR, SSIM, and MSE numbers than conventional mean and Wiener filters. Improved visual clarity is the result of the Kalman filter’s greater decrease in noise skills and preservation of the underlying structure of the film’s pixels. As a result, the precision of current traffic object recognition systems is greatly improved by these excellent frames. Applying the Kalman filter produced noticeably better outcomes for each studied output parameter producing MSE of 0.000123, PSNR of 42.35 and SSIM of 0.998 respectively. The tool used for execution is python.

A critical task in the field of monitoring and traffic administration is the identification of vehicles in CCTV footage. The scope of the proposed work is vehicles detection from CCTV videos. This article provides a thorough analysis of three well-known feature retrieval methods for detecting vehicles in CCTV images: SURF (Speeded-Up Robust Features), HOG (Histogram of Oriented Gradients), and KAZE. RMSE (Root Mean Square Error), MSE(Mean Square Error), and Silhouette Score are some of the assessment measures used in this study. In terms of RMSE, MSE, and Silhouette Score, the research results show that KAZE operates better than SURF and HOG, proving that it’s better at detecting fine features and durability in a variety of lighting and settings. The novelty of proposed work is better at vehicle detecting fine features from CCTV videos. This research also emphasizes how crucial it is to use the right feature extraction methods for precise and effective vehicle identification in practical settings. Applying the KAZE produced noticeably better outcomes for each studied output parameter producing RMSE of 0.02709, MSE of 0.000115 and Silhouette Score of 0.2 respectively. The tool used for execution Jupyter Notebook and language used is python.

Cervical cancer is a major health concern, and healthcare images play a major role in the analysis and handling of this disease. Three popular deep learning models that can be applied for cervical tumor identification and segmentation. The 3D U-Net model is a customized version of the standard U-Net framework, designed to handle 3D medical imaging data. DeepLab v3+ is another popular semantic segmentation model that uses atrous convolution to confine multi-scale related data. RPN is a popular object recognition model that applies a deep CNN to propose candidate regions in an image that may contain an object of interest. The cervical Cancer Risk Classification Dataset is collected from UCI Repository for assessment of the suggested DL models. The outcome of the DL models is evaluated based on the Dice Similarity Coefficient (DSC), Hausdorff Distances (HD) and Kappa Score (KS). Among the three models 3D U-Net provides better outcomes based on measured output metrics and gave DSC of 0.996, KS of 0.820 and HD of 9.7526 respectively. The tool used for execution is Matlab.

Cloud has a revolutionary change in Information Technology (IT) for data storage and retrieval operations compared to the traditional system. The drastic change in demand for cloud services has put several challenges for efficient resource allocation to customers. Moreover, competitive cloud service delivery and Service Level Agreement (SLA) violation have required a proficient technique to manage cloud resources. But, traditional resource management policies are unable to provide an appropriate match, hence inappropriate match leads to performance degradation. Swarms are capable of efficiently identify resource requirements through the computation process by using the available number of Virtual Machines (VMs) and allowing their optimal utilization. This research work has opted Ant Colony Optimization (ACO). The new proposed Adaptive Resource Availability Based Multiple Ant Colony Optimization (RABMACO) algorithm has generated an optimal solution for VMs allocation based on availability. The research work addressed in the way for developing a method used to optimize the performance of existing cloud environment by taking parameters for ACO algorithm, which was further experimentally determined. Then, the ACO algorithm has been optimized to the next level by developing resource availability based VM configuring and allocation. The experiment has been implemented with Datacenter, Host and a set of 5–50 VMs for running 100–1000 tasks of Montage dataset under the work flow sim simulation platform. The results have been evaluated on the basis of execution cost, execution time and VMs utilization. It has improved the availability of resources by releasing VMs earlier for performing next set of tasks.

Ensuring the quality of a product is crucial in the business, and it involves conducting checks, implementing control measures, and monitoring the process. Timely identification of product flaws is vital in the realm of manufacturing quality control. The utilization of automatic defect-detection technology offers more benefits compared to the manual identification of flaws. The initial section of the paper introduces a comprehensive classification system for various types of defects, which may be categorized into six distinct groups: Stain, pitted surface, Crack, black spots, Line, and Mono weld flaw. These faults would lead to a rise in the cost of the product and a decrease in the service life of the manufactured goods. The subsequent section of this article outlines the current state of traditional techniques and learning-based approaches in defect identification within the manufacturing business. We proceed with an examination of several defect detection methods, including statistical, spectral, model-based, and learning-based approaches. The primary objective of this study is to categorize the imperfections found in various items, including fabric material, steel, metal components, leather products, beverage products, and ceramic tiles. A comprehensive analysis has been conducted to evaluate and compare various automated defect detection methods and algorithms based on their distinctive features, accuracy in detecting defects, as well as their strengths and weaknesses.