Browse Items (2150 total)
Sort by:
-
Comparison between Symmetrical and Asymmetrical 13 Level MLI with Minimal Switches
Voltage source converters that are dependable and of the highest quality are offered by Multilevel Inverter to convert DC power systems to the AC power grid. One of the intriguing technologies in the field of power electronics are multilevel inverters (MLIs) in various configurations. It is also possible to integrate a few DC sources in MLIs to create a singular output, reducing the number of isolated inverters, the overall component count, and losses. MLIs are the top converters in many applications because to their capacity for medium and high-power applications. In order to produce the levels for the stair case wave shape, this research work introduces a new configuration module for asymmetrical multilevel in which capacitors are employed as DC linkages. With two unequal DC sources, the suggested Box -type modular structure will produce more voltage levels. It is useful for a variety of renewable applications since it has two back-to-back T-type inverters and minimal parts. This module contains this structured method to lessen the Total Harmonic Distortion (THD) rating and raise the quality of the sinusoidal output voltage. 2022 IEEE -
Comparison of Affine and DCGAN-based Data Augmentation Techniques for Chest X-Ray Classification
Data augmentation, also called implicit regularization, is one of the popular strategies to improve the generalization capability of deep neural networks. It is crucial in situations where there is a scarcity of high-quality ground-truth data. Also getting new samples is expensive and time consuming. This is a typical issue in the medical domain. Therefore, this study compares the performance of Affine and Generative Adversarial Networks (GAN)- based data augmentation techniques on the chest image X-Ray dataset. The Pneumonia dataset contains 5863 chest X-Ray images. The traditional Affine data augmentation technique is applied as a pre-processing technique to various deep learning-based CNN models like VGG16, Inception V3, InceptionResNetV2, DenseNet-169 and DenseNet-202 to compare their performance. On the other hand, DCGAN architecture is applied to the dataset for augmentation. Evaluation measures like accuracy, recall and AUC depict that DCGAN outperforms other traditional models. The most important advantage of DCGAN is that it is able to identify fake images with 100% accuracy. This is especially relevant for the medical domain as it deals with the life of individuals. Thus, it can be concluded that DCGAN has better performance as compared to affine transformations applied to traditional CNN models. 2023 The Authors. Published by Elsevier B.V. -
Comparison of Augmentation and Preprocessing Techniques for Improved Generalization Performance in Deep Learning based Chest X-Ray Classification
Convolutional Neural Network (CNN) models are well known for image classification; however, the downside of CNN is the ineffectiveness to generalize and inclination towards over-fitting in case of a small train dataset. A balanced and sufficient data is thus essential to effectively train a CNN model, but this is not always possible, especially in the case of medical imaging data, as often patients with the same disease are not always available. Image augmentation addresses the given issue by creating new data points artificially with slight modifications. This study, investigates ten different methods with various parameters and probability and their combined effect on the test dataset's generalization performance and F1 Score. For the study, three pre-Trained CNN models, namely ResNetl8, ResNet34, and ResNet50, are fine tuned on a small training dataset of 500 Pneumonia and 160 Non-Pneumonia(Normal) Images for each augmentation setting. The test accuracy, F1 Score, and generalization performance were calculated for a test dataset consisting of 50 Pneumonia and 16 Non-Pneumonia(Normal) Images. 2022 IEEE. -
Comparison of DQ Method with I cos? Controller in Solar Power System Connected to Grid with EV Load
Electric Vehicles and Photovoltaic power generation integrated to grid introduces power quality issues. Power quality issues during power integration needs improvement. Control of grid interfaced converters improves grid side power quality in integrated solutions. Power injection to the grid is controlled to get rid of power quality issues. Control techniques that can improve the power injection to the grid needs to be analyzed. This paper compares DQ and I cos ? controller while PV and EVs with non-linear loads are also connected in the power grid. Performance evaluation of both controllers are analyzed by comparing power injection to the grid. 2022 IEEE. -
Comparison of Full Training and Transfer Learning in Deep Learning for Image Classification
The deep learning algorithms on a small dataset are often not efficient for image classification problems. Make use of the features learned by a model trained on large similar dataset and saved for future reference is a method to solve this problem. In this work, we present a comparison of full training and transfer learning for image classification using Deep Learning. Three different deep learning architectures namely MobileNetV2, InceptionV3 and VGG16 were used for this experiment. Transfer learning showed higher accuracy and less loss than full-training. According to transfer learning results, MobileNetV2 model achieved 98.96%, InceptionV3 model achieved 98.44% and VGG16 model achieved 97.405 as highest test accuracies. The full-trained models did not achieve as much accuracy as that of transfer learning models on the same dataset. The accuracies achieved by full-training for MobileNetV2, InceptionV3 and VGG16 are 79.08%, 73.44% and 75.62% respectively. 2021, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. -
Comparison of Gradient Boosting and Extreme Boosting Ensemble Methods for Webpage Classification
Web page classification is an important task in various areas like web content filtering, contextual advertising and maintaining or expanding web directories etc. Machine Learning methods have been found to perform well to classify web pages, and ensemble models have been used to improve the results obtained from single classifiers. The Gradient Boosting and Extreme Boosting ensemble models are used in this work for binary classification. The dataset containing URLs of web pages have been collected manually. The comparison between the two boosting algorithms validated the improvement in accuracy and speed obtained through Extreme boosting. Extreme boosting has been found to be around ten times faster than Gradient boosting and also shows improvement in accuracy. The effect of three preprocessing techniques; lemmatization, stop words removal and regular expressions shows that these preprocessing techniques improves the accuracy of the results but not significantly. 2020 IEEE. -
Comparison of HOG and fisherfaces based face recognition system using MATLAB
Face recognition and validation is not an easy task due to barriers in between like variation in pose, facial expressions and illumination. There are many algorithms available to build a face recognition system. One such popular method of approach is the Histogram of Oriented Gradients (HOG). It is a simple but effective algorithm. Even though it gives satisfactory results, it sometimes mismatches query image with irrelevant images, especially in poor lighting conditions. This paper presents a more accurate technique called Fisherfaces. It is a more reliable method for face recognition and validation. Fisherface algorithm is utilized primarily for reducing the dimensionality of the feature space. Fisherface method for image recognition involves a series of steps. Firstly, the face space dimension is reduced using Principal Component Analysis (PCA) method, then the Linear Discriminant Analysis (LDA) method is used for feature extraction. Fisherface method produced good results even under complex situations like varying illumination conditions and images with different poses and expressions which is a major drawback of HOG. Fisherface algorithm can reach a maximum accuracy of 96.87%. Error Correcting Output Code (ECOC) is the classifier used for both HOG and Fisherfaces. 2021 IEEE. -
Comparison of Machine Learning Algorithms for Predicting Chronic Kidney Disease
Early detection and characterization of chronic renal disease are crucial to ensure that patients receive the best possible treatment. This study uses data mining techniques to uncover hidden information about patients. The outcomes of using the Random Forest, Multilayer Perceptron (MLP), Support Vector Machine (SVM), Decision Tree, XGBoost, LGBM Classifier, GaussianNB, KNeighbors Classifier, and XGBRF classifier have been compared. In our study, we demonstrate that Random Forest and XGBoost algorithms are more effective in classifying and predicting the severity level of chronic kidney disease 2022 IEEE. -
Comparison of Machine Learning-Based Intrusion Detection Systems Using UNSW-NB15 Dataset
Various machine learning classifiers have been employed recently to enhance network intrusion detection. In the literature, researchers have put forth a wide range of intrusion detection solutions. The accuracy of the machine learning classifiers intrusion detection is limited by the fact that they were trained on dated samples. Therefore, the most recent dataset must be used to train the machine learning classifiers. In this study, UNSW-NB15, machine learning classifiers are trained using the most recent dataset. A taxonomy of classifiers based on eager and lazy learners is used to train the chosen classifiers, such as K-Means (KNN), Polynomial Features, Random Forest (RF), and Naive Bayes (NB), Linear Regression. In order to decrease the redundant and unnecessary features in the UNSW-NB15 dataset, chi-Square, a filter-based feature selection technique, is used in this study. When comparing these machine learning classifiers, performance is measured in terms of accuracy, mean squared error (MSE), precision, recall, and F1-score with or without feature selection technique. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd 2024. -
Comparison of Various Types of Lubrication During Hard Turning of H13 Tool Steel by Analysing Flank Wear Using ANOVA
Hard machining of components has been a new attraction in the field of manufacturing, as it avoids the need for multiple cost inculcation processes for a single part. Hard machining attracts a wide attention to the researchers because of the usage of hard tools, tougher machinery and enormous quantities of cutting fluids. Optimized use of any of these functionaries can result in reduction of cost as well as safer and clean working environments. In this research new cutting fluid reduction processes were compared along with the use of hard metal inserts. These two methods suggest an enormous amount of cost reduction along with cleaner shop floor. Minimal quantity lubrication (MQL) and minimal cutting fluid application (MCFA) capacities in cutting fluid reduction as mentioned by various researchers in past two decades. These methods were compared in this research paper for finding out the best possible system. Flank wear is considered as a crucial parameter in hard machining as the wear rate affects other deserving product qualities such as surface finish and job profiles. In this research tungsten carbide coated hard metal inserts were used instead of conventional CBN or diamond tipped tools, which are of higher in price margin. The study comprised of Taguchis L9 orthogonal array, which was advised by previous researchers as good tool for optimisation. MQL and MCFA assisted experimentation were performed with same cutting conditions, which were then again compared with dry hard machining and wet machining. Influence of each input parameters where critically evaluated using ANOVA. The results revealed that a promising reduction in tool wear was noticed in MCFA assisted hard machining. 2020, Springer Nature Singapore Pte Ltd. -
Comparisons of Stock Price Predictions Using Stacked RNN-LSTM
This paper seeks to identify how the RNN-LSTM can be used in predicting the rise and fall in stock markets thereby helping investors to understand stock prices. Therefore, by predicting the nature of the stock market, investors can use different machine learning techniques to understand the process of selecting the appropriate stock and enhance the return investments thereafter. Long Short-Term Memory (LSTM) is a deep learning technique that helps to analyze and predict the data with respect to the challenges, profits, investments and future performance of the stock markets. The research focuses on how neural networks can be employed to understand price changes, interest patterns and trades in the stock market sector.The datasets of companies such as IBM, Cisco, Microsoft, Tesla and GE were used to build the stacked RNN-LSTM model using timesteps of 7 and 14days. The two layered stacked RNN-LSTM models of the companies such as Microsoft and Tesla achieved their highest model accuracies after being trained over a span of one year whereas the other companies acquired their highest accuracies after being trained over a span of 4 to 5years which implies that the rate of change of economic factors affecting Microsoft and Tesla over a short span of time is high as compared to the other existing companies. 2021, Springer Nature Switzerland AG. -
Compendium of Qubit Technologies inQuantum Computing
Quantum computing is information processing based on the principles of quantum mechanics. Qubits are at the core of quantum computing. A qubit is a quantum state where information can be encoded, processed, and readout. Any particle, sub-particle, or quasi-particle having a quantum phenomenon is a possible qubit candidate. Ascendancy in algorithms and coding demands knowledge of the specificities of the inherent hardware. This paper envisages qubits from an information processing perspective and analyses core qubit technologies. 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. -
Comprehensive Comparative Analysis of Breast Cancer Forecasting Using Machine Learning Algorithms and Feature Selection Methods
This research leveraged machine learning models, including Deep Neural Network (DNN), Artificial Neural Network (ANN), and Support Vector Machine (SVM), to predict breast cancer from CT and MRI scans. A dataset comprising 2345 instances of malignant and benign cases was meticulously curated, with 80% allocated for training and 20% for testing. The experimental results revealed the DNN as the top-performing model, exhibiting remarkable accuracy (95.2%), precision (94.8%), recall (95.6%), and F1 score (95.2%). The ANN also demonstrated strong performance, achieving an accuracy of 93.6% with balanced precision and recall scores. In contrast, the SVM, while respectable, fell slightly behind the machine learning models in terms of overall accuracy and performance. Detailed confusion matrices further elucidated the models capabilities and limitations, providing valuable insights into their diagnostic prowess. These findings hold great promise for breast cancer diagnosis, offering a non-invasive and highly accurate means of early detection. Such a tool has the potential to enhance patient care, reduce the strain on healthcare systems, and alleviate patient anxiety. The success of this research highlights the transformative impact of advanced machine learning in medical imaging and diagnosis, signaling a path toward more efficient and effective healthcare solutions. Further research and clinical validation are essential to translate these promising results into practical applications that can positively impact patients and healthcare providers. The Author(s), under exclusive license to Springer Nature Switzerland AG 2024. -
Comprehensive Review on Video Watermarking Security Threats, Challenges and its Applications
Data is a crucial resource for every business, and it must be protected both during storage and transmission. One efficient way of securing data and transferring it is through digital watermarking, where data is hidden inside a medium like text, audio, or video. Video watermarking is visible or invisible embedded data on a video in a logo, text, or video copyright disclaimer. In this proposed paper, the goal is to analyze the characteristics of video watermarking algorithms and the different metrics used for them. It deals with the extent to which the different requirements can be fulfilled, taking into consideration the conflicts between them and the practical challenges of video watermarking in terms of attacks like geometric attacks and non-geometric attacks. It also focuses on the process of watermarking a video. Recent advances in data security indicate that employing a video watermarking technology to transmit private data will be an effective method of transmitting sensitive data. The Electrochemical Society -
Comprehensive Study on Sentiment Analysis: Types, Approaches, Recent Applications, Tools and APIs
Sentiment analysis can be considered a major application of machine learning, more particularly natural language processing (NLP). As there are varieties of applications, Sentiment analysis has gained a lot of attention and is one among the fastest growing research area in computer science. It is a type of data analysis which is observed from news reports, user reviews, feedbacks, social media updates etc. Responses are collected and analyzed by researchers. All sentiments can be classified into three categories-Positive, Negative and Neutral. The paper gives a detailed study of sentiment analysis. It explains the basics of sentiment analysis, its types, and different approaches of sentiment analysis. The recent tools and APIs along with various real world applications of sentiment analysis in various areas are also described briefly. 2020 IEEE. -
Compression Based Modeling for Classification of Text Documents
Classification of text data one of the well known, interesting research topic in computer science and knowledge engineering. This research article, address the classification of text files issue using lzw text compression algorithms. LZW is a lossless compression technique which requires two pass on the input data. These two passes are treated separately as training stage and text stage for classification of text data. The proposed compression based classification technique is tested on publically available datasets. Results of the experiments shows the effectiveness of the proposed algorithm. 2019, Springer Nature Singapore Pte Ltd. -
Computational approach of artificial neural network
This paper makes an attempt to predict the movement of the stock price for the following day using Artificial Neural Network (ANN). For the purpose of this research, two companies from each industry have been chosen that is, TATA Motors and Honda Motors from the Automobile industry and Cadila Pharmaceuticals Ltd. and Glenmark Pharmaceuticals from the Pharmaceutical industry. The historical prices of these companies were collected and by using Artificial Neural Network (ANN), the movement of the stock price for the next day is predicted. 2017 IEEE. -
Computational Methods to Predict Suicide Ideation among Adolescents
Suicide has been a prominent cause of death worldwide, regardless of age, sex, geography, and so on, and predominantly suicide among teens, increased as the years have passed. Suicide ideation, suicide risk, suicide attempts have been studied extensively, and the most common cause has been identified as depression, followed by familial concerns, hereditary factors, stress, avoidance fear, and a variety of other variables. When visited by a doctor, most adolescents are unaware of their mental state and hence do not take action on their own or are not assisted by family or peer members to overcome their fear of social stigma or the treatment they must undergo. According to popular belief, early treatment and detection are the most effective ways to reduce the risk of suicide. As a result, the focus of this study is to illustrate some of the computational strategies utilized in deep learning and machine learning fields to detect kids at risk of suicide 2022 IEEE. -
Computational Modelling of Complex Systems for Democratizing Higher Education: A Tutorial on SAR Simulation
Engineering systems like Synthetic Aperture Radar (SAR) are complex systems and require multi-domain knowledge to understand. Teaching and learning SAR processing is intensive in terms of time and resources. It also requires software tools and computational power for preprocessing and image analysis. Extensive literature exists on computational models of SAR in MATLAB and other commercial platforms. Availability of computational models in open-source reproducible platforms like Python kernel in Jupyter notebooks running on Google Colaboratory democratizes such difficult topics and facilitates student learning. The model, discussed here, generates SAR data for a point scatterer using SAR geometry, antenna pattern, and range equation and processes the data in range and azimuth with an aim to generate SAR image. The model demonstrates the generation of synthetic aperture and the echo signal qualities as also how the pulse-to-pulse fluctuating range of a target requires resampling to align the energy with a regular grid. The model allows for changing parameters to alter for resolution, squint, geometry, radar elements such as antenna dimensions, and other factors. A successful learning outcome would be to understand where parameters need to be changed, to affect the model in a specific way. Factors affecting Range Doppler processing are demonstrated. Use of the discussed model nullifies use of commercial software and democratizes SAR topic in higher education. 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. -
Computationally Efficient Machine Learning Methodology for Indian Nobel Laureate Classification
A computationally efficient methodology for Indian Nobel Laureate classification is proposed in this study, emphasizing the optimization of image categorization through supervised learning techniques. Leveraging advancements in Convolutional Neural Networks (CNNs), the research aims to enhance the efficiency and precision of image classification tasks. The study utilizes Logistic Regression for dataset analysis, initially employing browser extensions for mass downloading categorized image data. Haar cascade classifiers are then used for data wrangling, focusing on facial, nose, and mouth recognition. Following this, feature engineering through wavelet transformation reduces image dimensionality, preparing the dataset for the chosen ML model, Logistic Regression. The primary focus is to simplify technology for improved image categorization. Support Vector Machines (SVM), Random Forest, and Logistic Regression are examined, with Logistic Regression emerging as the most effective model, achieving an accuracy rate of 87.5%. A thorough evaluation using Confusion Matrices reveals Logistic Regression's superior performance in classifying images of Indian Nobel laureates. A strategic up-sampling approach is implemented to address dataset inconsistencies, ensuring balanced representation across classes. The Haar wavelet transform is then applied for feature extraction, optimizing the dataset for ML models. The dataset is split into training and testing sets (80-20), and the three models are trained and evaluated for accuracy. Logistic Regression proves to be the best performer, offering insights into prominent leaders' identification. The research offers a detailed pipeline for data preprocessing, feature engineering, and model assessment, culminating in a robust image categorization system. Logistic Regression emerges as a reliable method for biographical picture identification, demonstrating superior accuracy over SVM and Random Forest. This research underscores the importance of efficient and accurate image classification methodologies for practical applications in real-world scenarios, particularly in recognizing influential leaders. 2024 IEEE.