International Journal on Recent and Innovation Trends in Computing and Communication
Not a member yet
8613 research outputs found
Sort by
Domain Classification for Marathi Blog Articles using Deep Learning
Get PDFNowadays the exponential growth of online content, particularly in the form of blog articles is tremendous, the need for effective techniques to automatically categorize them into relevant domains has become increasingly important. To overcome the challenges the domains like natural language processing (NLP), machine learning (ML) and deep learning (DL)are being working as booster effect to emerge out with solutions. In this proposed system methodology-based NLP and DL domain the long short-term memory (LSTM) classifier for domain classification and compared the existing multiclass classification techniques with having accuracy around 94% and 91% by long short-term memory (LSTM) model using two different data sets one is Marathi new article and another one Financial article data set. The proposed model is being compared with multiple other models like naïve bayes (NB), XGBoost, support vector machine (SVM) and random forest (RF). The final estimated result achieved is best combination of dataset and deep learning algorithm LSTM
Fusion based Image Enhancement Approach for Brain Tumor Detection
Get PDFMagnetic Resonance Imaging (MRI), is a crucial technology used in the processing of medical images that provides insights into the anatomy of soft organs in the human body and helps in detecting brain tumors and spinal tumors. Despite advances in technology, most images have intrinsic drawbacks such as reduced contrast and brightness, and noise. Several contrast enhancement techniques are used such as, HE, BBHE, DSIHE, CLAHE, RMSHE, and their fusion, have been deployed on different MRI images to handle these problems. Metrics such as, entropy, PIQE and BRISQUE are used in the assessment of the results. Through the different fusion combinations, most prominent results are obtained from CLAHE-RMSHE fusion with an entropy value of 6.2516 and BRISQUE value of 40.14
A Hybrid Resampling Approach for Multiclass Skewed Datasets and Experimental Analysis with Diverse Classifier Models
Get PDFIn real-life scenarios, imbalanced datasets pose a prevalent challenge for classification tasks, where certain classes are heavily underrepresented compared to others. To combat this issue, this article introduces DOSAKU, a novel hybrid resampling technique that combines the strengths of DOSMOTE and AKCUS algorithms. By integrating both oversampling and undersampling methods, DOSAKU significantly reduces the imbalance ratio of datasets, enhancing the performance of classifiers. The proposed approach is evaluated on multiple models employing different classifiers, and the results demonstrate its superiority over existing resampling measures, making it an effective solution for handling class imbalance challenges. DOSAKU's promising performance is a substantial contribution to the field of imbalanced data classification, as it offers a robust and innovative solution for improving predictive model accuracy and fairness in real-world applications where imbalanced datasets are common
Imputation Techniques in Machine Learning – A Survey
Get PDFMachine learning plays a pivotal role in data analysis and information extraction. However, one common challenge encountered in this process is dealing with missing values. Missing data can find its way into datasets for a variety of reasons. It can result from errors during data collection and management, intentional omissions, or even human errors. It's important to note that most machine learning models are not designed to handle missing values directly. Consequently, it becomes essential to perform data imputation before feeding the data into a machine learning model. Multiple techniques are available for imputing missing values, and the choice of technique should be made judiciously, considering various parameters. An inappropriate choice can disrupt the overall distribution of data values and subsequently impact the model's performance. In this paper, various imputation methods, including Mean, Median, K-nearest neighbors (KNN)-based imputation, Linear Regression, Miss Forest, and MICE are examined
Network Intrusion Detection Using Autoencode Neural Network
Get PDFIn today's interconnected digital landscape, safeguarding computer networks against unauthorized access and cyber threats is of paramount importance. NIDS play a crucial role in identifying and mitigating potential security breaches. This research paper explores the application of autoencoder neural networks, a subset of deep learning techniques, in the realm of Network Intrusion Detection.Autoencoder neural networks are known for their ability to learn and represent data in a compressed, low-dimensional form. This study investigates their potential in modeling network traffic patterns and identifying anomalous activities. By training autoencoder networks on both normal and malicious network traffic data, we aim to create effective intrusion detection models that can distinguish between benign and malicious network behavior.The paper provides an in-depth analysis of the architecture and training methodologies of autoencoder neural networks for intrusion detection. It also explores various data preprocessing techniques and feature engineering approaches to enhance the model's performance. Additionally, the research evaluates the robustness and scalability of autoencoder-based NIDS in real-world network environments. Furthermore, ethical considerations in network intrusion detection, including privacy concerns and false positive rates, are discussed. It addresses the need for a balanced approach that ensures network security while respecting user privacy and minimizing disruptions. operation. This approach compresses the majority samples & increases the minority sample count in tough samples so that the IDS can achieve greater classification accuracy
Network Based Intrusion Detection System Using Weighted Product Model (WPM)
Get PDFA security technology called a network-based intrusion detection system (NIDS) was created to safeguard computer networks against unauthorised access and criminal activity. This technology works by analysing network traffic, spotting potential risks, and informing administrators of any possible incursions or attacks. NIDS research ensures that intrusion detection systems are built to minimise the gathering and storage of sensitive data by taking into account the value of privacy and data protection .In general, network-based intrusion detection system research has a major impact on how well these security measures operate, how efficiently they perform, and how adaptable they are.By addressing the evolving challenges posed by cyber threats, NIDS research helps organizations enhance their network security posture, protect sensitive information, and defend against potential intrusions and attacks." The weighted product model (WPM), a multi-criteria decision-making (MCDM) technique, is used to evaluate and rank solutions based on a variety of distinct criteria. It provides a methodical approach to decision-making by considering the relative importance of each attribute and the performance of other solutions in relation to those criteria. The WPM normalises the data, weights the criteria, and gives a weighted score for each alternative. The option with the greatest score is regarded as the ideal option. The weighted product model offers a structured framework for making decisions by taking into account many factors and their varying degrees of importance. It enables decision-makers to assess and contrast options using a wide range of criteria, resulting in more informed and unbiased choices. It's crucial to check nonetheless that the model's weights and normalisation techniques appropriately capture the decision-maker's preferences as well as the features of the choice problem.J48, Random Forest, JRIP, RIDOR, PART. The definition of true positive, false positive, true negative and false negative rates has already been established. These metrics for measuring the effectiveness of classification algorithms, anomaly detection systems, and binary decision-making processes are accurately presented. As can be seen from the results, J48 received the highest rank, while PART received the lowest .In order to increase the security of computer networks, network-based intrusion detection systems (NIDS) are essential. They provide real-time monitoring and analysis of network traffic to identify suspected breaches and malicious activities, enabling appropriate action to be taken. However, it is important to recognize that NIDS can have limitations and are not infallible
Big Data Network Optimization for Mobile Cellular Networks in 5G
Get PDF5G ensures the provision of intelligent network and application services by means of connectivity to remote sensors, massive amounts of Internet of Things data, and fast data transmissions. Through the utilization of distributed compute architectures and by supporting massive connectivity across diverse devices like sensors, gateways, and controllers, 5G brings about a transformative revolution in the conversion of both big data at rest and data in motion into real-time intelligence. Big Data Analytics play an important role in the evolution of 5G standards, enabling intelligence across networks, applications, and businesses. Administrators of mobile organizations have access to a plethora of opportunities to enhance service quality through big data. Network optimization serves as a crucial method to achieve this task, with network prediction forming the foundation for such optimization. Ensuring network stability and security is essential for 5G mobile communication, considering its significance as an important tool in national life. Therefore, this work focuses on presenting big data network optimization for mobile cellular networks within the context of 5G. In order to improve the Quality of Experience (QoE) for users, this work explores various methods for integrating network optimization and Big Data analytics. The performance of the presented model is evaluated in terms of QoE, Throughput, handover rate, mobility, reliability, and network slicing
Automating Risk Assessment: The Role of Artificial Intelligence in Insurance Underwriting
Get PDFAutomating risk assessment through artificial intelligence (AI) can significantly transform the insurance underwriting process by improving accuracy and efficiency. This paper explores the development and implementation of a hybrid machine learning model that integrates logistic regression and support vector machines (SVM) for enhanced underwriting risk assessment. By meticulously analyzing historical claim data and detailed customer profiles, this hybrid model achieves a notable accuracy of 91%, far surpassing the 81.3% accuracy typically associated with manual underwriting practices. Logistic regression is utilized for its simplicity and effectiveness in modeling relationships between dependent and independent variables. It helps in identifying key risk factors from the dataset, providing clear insights into how various customer attributes influence claim likelihood. Support vector machines (SVM) are then applied to classify and predict the likelihood of claims, leveraging their strength in handling both linear and non-linear data. The combination of these two methods results in a robust predictive model capable of delivering highly accurate risk assessments. The model's ability to predict claims with such high accuracy not only enhances the precision of risk assessments but also significantly speeds up the underwriting process. This reduction in processing time can lead to faster policy issuance, improved customer satisfaction, and operational efficiencies within insurance companies. Furthermore, the AI-driven approach enables insurers to identify high-risk individuals more accurately, allowing for better resource allocation and risk management
Predictive Maintenance in Manufacturing: Utilizing Machine Learning for Equipment Health Monitoring
Get PDFPredictive maintenance utilizing machine learning is crucial for optimizing equipment health in manufacturing environments. This research presents a novel hybrid machine learning model that combines Support Vector Machines (SVM) and Recurrent Neural Networks (RNN) to predict equipment failures accurately. The model is built on a foundation of systematically collected sensor data and operational metrics, which undergo extensive preprocessing to ensure data quality and integrity. The SVM component is adept at classifying current equipment health states, while the RNN, particularly its Long Short-Term Memory (LSTM) networks, excels in analyzing temporal sequences of sensor data to predict future equipment conditions. This dual approach enables the model to achieve a high prediction accuracy of 91.4%. The implementation of this predictive maintenance model in a manufacturing plant has yielded significant operational benefits. Specifically, the model's real-time monitoring and alert system facilitated a 25% reduction in equipment downtime. Moreover, by enabling timely and accurate maintenance interventions, the model contributed to a 15% decrease in maintenance costs. The architecture of the developed system is robust and comprehensive, encompassing real-time data acquisition from IoT sensors, centralized data storage, and rigorous data processing. The continuous monitoring feature ensures that maintenance personnel are promptly alerted to potential issues, allowing for proactive measures that prevent equipment failures and minimize unplanned downtime. These results highlight the effectiveness of the hybrid SVM-RNN model in enhancing the reliability and efficiency of manufacturing operations. By leveraging advanced machine learning techniques, this predictive maintenance strategy demonstrates significant improvements in operational performance and cost savings. This study underscores the potential of integrating machine learning into maintenance practices to achieve greater precision and efficiency in manufacturing settings
Control Techniques for MPPT of Grid Connected Photovoltaic Systems
Get PDFA grid-connected PV system is analyzed in this paper using MPPT techniques. A major goal of research has been finding methods that extract solar energy more efficiently, which has attracted numerous researchers towards photovoltaics (PVs). The nonlinear behavior of the device determines its output based on a variety of factors, including solar temperature and ambient irradiance. Using MPPT techniques, PV cells can achieve their maximum output. P&O and incremental conductance are two MPPT techniques that are most commonly used in this paper. As well as PWM techniques, three grid-connected inverters operating at 240 kV are compared. Solar energy systems must use Maximum Power Point Tracking (MPPT) to maximize power output under variable irradiation and climate conditions. This research deals with two MPPT algorithms for PV arrays connected to the grid: Perturb and Observe (P&O), and Fuzzy Logic Control (FLC). Analyzing and comparing the results of both algorithms is performed using MATLAB/Simulink