Statistics, Optimization & Information Computing

Inferential study on lifetime performance index with generalized inverted exponential model under progressive first-failure censoring

Huijun Yi, Danush K. Wijekularathna — Thu, 31 Oct 2024 00:00:00 +0800

Lifetime performance assessment is widely used in quality control of the manufacturing industry. This paper focuses on the progressively first-failure-censored data coming from the generalized inverted exponential distribution. We present the maximum likelihood estimate and the Bayesian estimate for the lifetime performance index (C_L) for a given lower specification level L. The results are used to develop non-Bayesian and Bayesian inferences to determine whether the product performance meets the required level. A Monte Carlo simulation and two real data examples are discussed for illustration purposes.

An Extended Discrete Model for Actuarial Data and Value at Risk Analysis: Properties, Applications and Risk Analysis under Financial Automobile Claims Data

Thu, 05 Dec 2024 00:00:00 +0800

This paper deals with a new discrete distribution with high flexibility. We have studied many of its mathematical and statistical properties, and we have neglected many other properties due to the narrow scope of the paper.
Additionally, we have presented a comprehensive analysis of actuarial risks. A good set of actuarial risk indicators that are used in financial analysis and measurement and evaluation of financial risks. Five discrete data sets have been relied upon in conducting the financial analysis and risk assessment. Necessary comments have been provided on the results of the paper, and a set of necessary recommendations are provided for insurance companies to avoid the occurrence of unexpected large losses. All these financial analyses have been conducted in light of a discrete probability distribution.

A New Generalization of the Inverted Gompertz Distribution with Different Methods of Estimation and Applications

Ibtesam Alsaggaf, Sara F. Aloufi — Tue, 20 Aug 2024 00:00:00 +0800

Designing appropriate models for analyzing data in various fields is essential as it helps professionals comprehend complex data patterns and their characteristics, leading to informed decision-making. Despite the diversity of probability distribution, the data may not conform to classical distributions in many instances. Consequently, there arises a need for a new distribution that can accommodate the intricacies of diverse data forms and enhance the goodness of fit. This article introduces a novel extended lifetime model called the new exponential exponentiated generalized inverted Gompertz based on the new exponential-X family of distributions. The article discusses some statistical properties associated with the proposed distribution. The parameters of the new distribution are estimated using multiple estimation techniques, and their performance is compared through Monte Carlo simulations. The demonstrated potential and effectiveness of the proposed distribution are exemplified by its application to three datasets within various fields.

Distance k-domination and k-resolving domination of the corona product of graphs

Dwi Agustin Retnowardani, Liliek Susilowati, Dafik, Kamal Dliou — Mon, 26 Aug 2024 00:00:00 +0800

For two simple graphs $G$ and $H$, the corona product of $G$ and $H$ is the graph obtained by adding a copy of $H$ for every vertex of $G$ and joining each vertex of $G$ to its corresponding copy of $H$. For $k \geq 1$, a set of vertices $D$ in a graph $G$ is a distance $k$-dominating set if any vertex in $G$ is at a distance less or equal to $k$ from some vertex in $D$. The minimum cardinality overall distance $k$-dominating sets of $G$ is the distance $k$-domination number, denoted by $\gamma_k(G)$. The metric dimension of a graph is the smallest number of vertices required to distinguish all other vertices based on distances uniquely. The distance $k$-resolving domination in graphs combines distance $k$-domination and the metric dimension of graphs. In this paper, we investigate for all $k\geq 1$, the distance $k$-domination and the distance $k$-resolving domination in the corona product of graphs. First, we show that for $k\geq 2$ the distance $k$-domination number of $G\odot H$ is equal to $\gamma_{k-1}(G)$ for any two graphs $G$ and $H$. Then, we give the exact value of $\gamma_{k}(G\odot H)$ when $G$ is a complete graph, complete $m$-partite graph, path and cycle. We also provide general bounds for $\gamma_{k}(G\odot H)$. Then, we examine the distance $k$-resolving domination number for $G\odot H$. For $k=1$, we give bounds for $\gamma^r(G\odot H)$ the resolving domination number of $G\odot H$ and characterize the graphs achieving those bounds. Later, for $k\geq 2$, we establish bounds for $\gamma^r_k(G\odot H)$ the distance $k$-resolving domination number of $G\odot H$ and characterize the graphs achieving these bounds.

Topp-Leone Type I Heavy-Tailed - G Power Series Class of Distributions: Properties, Risk Measures, and Applications

Wilbert Nkomo, Broderick Oluyede, Fastel Chipepa — Thu, 15 Aug 2024 00:00:00 +0800

This study presents a new class of distributions (CoDs) called the Topp-Leone type I heavy-tailed-G power series (TL-HT-GPS),
along with its subclass, the Topp-Leone type I heavy-tailed log-logistic power series (TL-HT-LLOGPS) distribution. 
Statistical properties of this novel CoDs were derived, and actuarial risk measures were developed and numerically simulated.
The maximum likelihood estimation technique was employed to estimate the unknown parameters of the model, and Monte Carlo 
simulations were used to evaluate the estimates' consistency. Through the use of the Topp-Leone type I heavy-tailed 
log-logistic Poisson (TL-HT-LLOGP) distribution, a special case of the TL-HT-LLOGPS distribution, two real data sets 
including a censored case, were examined to illustrate the potential of the proposed distribution. The TL-HT-LLOGP 
distribution was compared to a few selected non-nested competing distributions including some known heavy-tailed 
distributions and power series distributions. The TL-HT-LLOGP out-performed the contending distributions through various
 goodness-of-fit tests conducted.

Extreme Value Stable Mixture Modelling with applications to South African stock market indices and exchange rate

Kimera Naradh, Knowledge Chinhamu, Retius Chifurira — Thu, 08 Aug 2024 00:00:00 +0800

In recent times, there is a vested interest in the research and applications of extreme value mixture models in the stock market and insurance as well as medical industries. This study aims to evaluate the fit of two extreme value mixture models namely Stable-Normal-Stable (SNS) and Stable-KDE-Stable (SKS), where KDE represents the Kernel density estimator, for three FTSE/JSE indices namely All Share Index (ALSI), Banks Index, Mining Index and the USD/ZAR currency exchange rate. These novel models aim to capture the characteristics of South African financial data as compared to the existing commonly fitted extreme value mixture models. The results highlight the robustness of the SNS and SKS mixed model for each daily returns when conducting a graphical bulk model and comprehensive tail model analysis. Financial practitioners looking to curb losses and study alternatives for financial modeling in the South African financial industry using an extreme value mixed model approach may find the SNS and SKS model application beneficial.

Time ‎‎truncated double acceptance sampling plan for the Nadarajah-Haghighi distribution

Mehran Naghizadeh Qomi, AL-Husseini Zainalabideen, Sanku Dey — Sun, 04 Aug 2024 00:00:00 +0800

In this article, we design a double acceptance sampling plan for the Nadarajah-Haghighi (NH) distribution when the lifetime is truncated. The minimum sample sizes necessary to ensure a certain mean lifetime for selected acceptance numbers and consumer's confidence levels are obtained. The operating characteristic function and the associated producer's risks are studied. We also analyze the minimum ratios of the mean life to the specified life. Real data and simulated examples are provided to illustrate the results of the paper.

Magnetic Resonance Image Restoration by Utilizing Fractional-Order Total Variation and Recursive Filtering

Nana Wei, Wei Xue, Xiaolei Gu, Xuan Qi — Tue, 10 Dec 2024 00:00:00 +0800

Total variation-based methods are effective for magnetic resonance image restoration. To eliminate impulse noise, the $\ell_0$-norm total variation model is a proven approach. However, traditional total variation image restoration often results in staircase artifacts, especially at high noise levels. In this paper, we propose an innovative magnetic resonance image restoration model that integrates fractional-order regularization and filtering techniques. Specifically, the first term uses the $\ell_0$-norm as the data fidelity term to effectively remove impulse noise. The second term introduces a fractional-order total variation regularizer, which preserves structural information while reducing staircase artifacts during deblurring. Due to its limitations in texture detail recovery, we employ recursive filtering for high-quality edge-preserving filtering. Finally, we solve the optimization model using the alternating direction method of multipliers. Experimental results demonstrate the effectiveness of our method in restoring magnetic resonance images.

Real-Time Scheduling Optimization of Integrated Energy Systems in Smart Grids based on Approximate Dynamic Programming

Sun, 17 Nov 2024 00:00:00 +0800

With the large-scale integration of renewable energy (RE) sources and rapid advancements in smart grid (SG) technologies, the efficient integration of diverse energy resources to achieve supply-demand balance and maximize costeffectiveness has emerged as a research hotspot in the energy sector. This paper addresses the real-time scheduling challenge in integrated energy systems (IES) within the context of SG, emphasizing pivotal factors such as electric and thermal load scheduling, energy storage control, dynamic electricity pricing, carbon emission mechanisms, and demand response (DR). To this end, we propose a comprehensive scheduling model tailored for IES, aiming to minimize the total cost over the dispatch cycle. Furthermore, an optimal scheduling algorithm based on approximate dynamic programming (ADP) was designed to solve this model. Numerical experiments reveal that, while ensuring user comfort, the proposed real-time scheduling scheme, by comprehensively considering the interactions among various system inputs, significantly enhances system flexibility and economic performance. It effectively tackles the uncertainty of RE, thereby improving energy utilization efficiency.

A Connection between the Adjoint Variables and Value Function for Differential Games

Rania Benmenni, Nourreddine Daili — Mon, 19 Aug 2024 00:00:00 +0800

In this paper, we present a deterministic two-player nonzero-sumd ifferential games (NZSDGs) in a finite horizon. The connection between the adjoint varaibles in the maximum principle (MP) and the value function in the dynamic programming principle (DPP) for differentail games is obtained in either case, whether the value function is smooth and nonsmooth. For the smooth case, the connection between the adjoint variables and the derivatives of the value function are equal to each other along optimal trajectories. Furthermore, for the nonsmooth case, this relation is represented in terms of the adjoint variables and the first-order super- and subdifferentials of the value function. We give an example to illustrate the theoretical results.

Parametric Support approach for solving Mean-Variance problem under general constraints

Souhaib Boudjelda, Belkacem Brahmi — Tue, 13 Aug 2024 00:00:00 +0800

The intuitive and natural formulation of the Mean-Variance (MV) model has attracted the attention of researchers over the years. This model is typically presented as a constrained Quadratic Problem (QP), although the practical aspects of investment often require risk tolerance to be considered. In such cases, Parametric Quadratic Programming (PQP) is employed to explore all optimal solutions on the efficient frontier. In this paper, we propose a novel approach for solving the portfolio optimization problem of the mean-variance model. This problem is considered in its parametric formulation under general linear equality constraints with bounded assets. The proposed algorithm iteratively derives the exact efficient frontier by calculating all corner portfolios as a function of the risk aversion parameter. Finally, we test the computational performance of our algorithm in comparison with two state-of-the-art approaches using a set of real benchmarks. The results demonstrate the effectiveness of our approach in solving such problems and in identifying the efficient frontier. Additionally, considering large-scale randomly generated problems with dense covariance matrices, we show that our algorithm can efficiently solve this class of problems in a reasonable computation time.

New Software Reliability Growth Model: Piratical Swarm Optimization -Based Parameter Estimation in Environments with Uncertainty and Dependent Failures

Fri, 23 Aug 2024 00:00:00 +0800

In this paper our software solutions are delivered and installed in field conditions that are either identical to or comparable to development and test environments. As a result, they may also be used in a variety of settings that differ from the ones in which they were created and tested. Software dependability can be hard to increase for a variety of reasons, including a particular environment or a flaw in the code. In this research, we offer a novel software reliability model that considers operating environment unpredictability. It has been explained the proposed model and other models of the non-homogeneous Poisson process (NHPP) is demonstrated with examples. Has been used two sets of defect data from software applications. We estimated all models’ parameters by using the Cuckoo Search algorithm (CS) technique. We also conducted a simulation process to determine the good model. Through the results and their comparison with other NHPP models used, the proposed model is better than the other models and fits the data better.

Advanced Big Data Analytics: Integrating Fuzzy C-Means, Encoder-Decoder CNNs, and Genetic Algorithms for Efficient Clustering and Classification

Fatima Belhabib, Mohamed BENSLIMANE, Karim El Moutaouakil — Sun, 15 Dec 2024 07:45:34 +0800

In the realm of Big Data analysis, the pivotal question of data clustering takes center stage. This study delves into optimizing this analysis by adopting a hybrid approach that integrates the Fuzzy C-Means (FCM) methodology, Encoder-Decoder Convolutional Neural Networks (CNN), Genetic Algorithms (GAs), and an optimal classification strategy for data clustering and categorization. FCM provides a flexible clustering foundation with its fuzzy logic, while the Encoder-Decoder CNN contributes to extracting complex features and refining the model. Genetic Algorithms finely adjust the parameters of the hybrid model. The optimal classification strategy complements this approach by ensuring precise data categorization. This hybrid strategy leverages the specific strengths of each component, thereby overcoming inherent limitations in each technique. FCM ensures robust cluster formation the Encoder-Decoder CNN improves feature representation, Genetic Algorithms optimize the hyper-parameters of the hybrid model, and optimal classification reinforces the accuracy of data categorization. Experiments conducted on various Big Data sets reveal a significant enhancement in clustering and classification accuracy, as well as overall analysis efficiency. This research represents a substantial contribution to the evolution of Big Data analysis by proposing an integrated solution harnessing the power of FCM, Encoder-Decoder CNN, Genetic Algorithms, and optimal classification The results suggest that this hybrid approach not only increases clustering and classification accuracy but also provides a versatile and adaptable solution to address challenges in large-scale data analysis.

Optimizing Kohonen Classification of Mixed Data with Partial Distance and Referent Vector Initialization

Mouad Touarsi, Driss Gretete, Abdelmajid Elouadi — Thu, 25 Jul 2024 00:00:00 +0800

The success of neural network models in clustering problems is highly dependent on the quality and diversity of the data used. Self-organizing maps (SOM), a semi-supervised data learning tool introduced by Kohonen in the 1980s, have been widely used in various fields such as signal and text recognition, industrial data analysis, speech and image recognition, etc. SOM's competitive learning clustering method, where each node specializes in a specific subset of data, has proven to be a powerful technique.

In this paper, we propose a new SOM variant suitable for handling numerical, interval, and categorical attributes simultaneously. Instead of random initialization of weights, we utilize the ASAICC algorithm to select initial referent vectors.

Furthermore, we suggest representing one cluster using multiple referent vectors at once. The effectiveness of the proposed Kohonen variant is evaluated using well-known benchmark datasets, and the results are reported using reliable performance metrics. The simulation of the new algorithm is conducted using the R language, and the obtained results demonstrate the superiority of the proposed approach.

Hybrid Approach for Minimizing Departure Air Traffic Delays Following Standard Instrument Departures

Abdelmounaime BIKIR, Otmane Idrissi, Khalifa Mansouri, Mohammed Qbadou — Sat, 24 Aug 2024 00:00:00 +0800

The efficient scheduling of departure air traffic persists as one of the most challenging aspects of air traffic management in recent years. A proper sequencing enhances airport operations, minimises delay, and improves airspace capacity and traffic forecasting. This paper proposes a sequential hybridisation algorithm designed to assist air traffic controllers in determining the optimal departure sequence complying with the standard instrument departures (SIDs).

The level of complexity increases when taking into account the departure runway constraints, the configuration of flight paths after takeoff, and the aircraft's operational limits during the takeoff phase. Another challenging aspect is the wide diversity in aircraft types.

The suggested approach proposes a Genetic algorithm (GA) strengthened with the Partially Mixed Crossover technique (PMX). The initial population of the GA is enhanced with the Shortest Job First (SJF) method. This sequential hybridisation algorithm dynamically arranges the departure traffic sequence based on their performances and the complexity of the followed SIDs.

A Generalized Backstepping Controller Design for a Second-Order Magnetic Levitation System

Oscar Danilo Montoya Giraldo, Walter Gil-González, Adolfo Jaramillo-Matta — Thu, 31 Oct 2024 00:00:00 +0800

This research tackles the control design challenge of stabilizing a second-order magnetic levitation system using a nonlinear control approach. The proposed controller is rooted in backstepping control theory, which ensures the asymptotic convergence of the system’s incremental state variables to the origin through a Lyapunov-based framework. A key advantage of this method is the generalized control input, expressed in a polynomial form with four adjustable control gains, allowing for precise tuning to achieve the desired dynamic performance. A major contribution of this study is the formal demonstration of stable performance provided by the generalized controller in second-order dynamic systems, with a particular emphasis on its application to magnetic levitation. Numerical simulations in Matlab/Simulink showcase the controller’s effectiveness across three different sets of control gains, enabling the system to realize critically damped, overdamped, and underdamped dynamic responses with respect to the desired position of the levitated metallic mass.

Detecting lung diseases from X-Ray images using deep learning

Bao Nguyen, Anh Vo H. — Wed, 09 Oct 2024 00:00:00 +0800

Lung disease has become one of the most dangerous diseases worldwide after the Covid-19 pandemic. Early diagnosis of lung disease is vital for effective treatment and recovery. In clinical practice, X-ray imaging is currently the most widely used method for diagnosis, and it plays a crucial role as a life-saving factor for individuals suffering from the disease. In recent years, many deep learning approaches have been proposed for the early diagnosis of lung diseases from X-ray images. These approaches have shown high accuracy in predicting the results within a short time. This paper aims to compare different state-of-the-art deep learning models for the task of lung-disease diagnosis. Additionally, we have collected a new dataset of lung disease X-ray images from hospitals in Vietnam to evaluate the performance of each model based on validation loss and validation accuracy. The results show that our proposed deep learning model achieves an accuracy of 98.35% (training) and 86.65% (validation) on the new ChestVN lung disease dataset, which promises to be a good method for applying in daily life. The proposed approach has the potential to assist medical professionals in the early diagnosis of lung diseases, which can lead to better patient outcomes and improved healthcare management.

Enhancing Performance and Latency Optimization in Fog Computing with a Smart Job Scheduling Approach

Meena Rani, Kalpna Guleria, Surya Narayan Panda — Wed, 09 Oct 2024 00:00:00 +0800

Nowadays, Internet of Everything (IoE) devices are growing rapidly, producing vast amounts of data. Cloud
computing offers processing, analysis, and storage solutions to manage these large data volumes. However, the rising latency and bandwidth usage could be more suitable for real-time applications, including intelligent healthcare devices, online gaming, and surveillance via video. In order to tackle the rise in latency and bandwidth utilization in cloud computing technology, Fog Computing (FC) has been developed as it offers networking, processing, storage, and analytics functions. Since FC is still in its early stage, scheduling jobs and allocating resources are two significant issues. With the help of this innovation, there are resource limitations on the fog devices at the network’s edge. Consequently, scheduling is crucial for choosing a fog node for a job assignment. An intelligent and effective work scheduling algorithm can decrease energy usage and application request response time. This research introduces an innovative Quality of Service Priority Tuple Scheduling (QoSPTS) scheduler that optimizes network capacity and latency while enabling service for the IoE. This case study demonstrates the effective management of IoE device requirements by efficiently allocating resources across fog devices and optimizing scheduling to enhance quality of life. The study uses iFogSim to compare the proposed scheduling algorithm with other methods by considering energy efficiency, network utilization, cost, and latency as performance measures. Results showed that the proposed Scheduler’s latency network bandwidth, energy utilization and cost are highly enhanced compared to traditional approaches such as FCFS, Concurrent, Delay-Priority, and QoS-Aware.

Sentiment Analysis in the Transformative Era of Machine Learning: A Comprehensive Review

Tue, 20 Aug 2024 00:00:00 +0800

Sentiment analysis, which stands for opinion mining, is a natural language processing (NLP) technique that involves identifying, extracting, and analyzing sentiments or opinions expressed in text data. The primary goal of sentiment analysis is to determine the sentiment polarity of a given piece of text, whether it is positive, negative, or neutral. This analysis can be applied to various types of content, such as product reviews, social media posts, customer feedback, and news articles. Sentiment analysis algorithms use machine learning and text classification to understand subjective information conveyed in text, helping businesses, organizations, and individuals gain insight into public opinions and emotions about specific topics, products, or services. In this study, we conducted sentiment analysis on a Bengali dataset. For feature extraction, we implemented the term frequency-inverse document frequency (TF-IDF) technique, and for feature selection, we applied an extra tree classifier approach. Subsequently, we trained our machine learning model, achieving an impressive accuracy rate of 92%.

Harnessing AI for Precision Oncology: Transformative Advances in Non-Small Cell Lung Cancer Treatment

Rihab EL SABROUTY, Abdelmajid ELOUADI — Fri, 02 Aug 2024 00:00:00 +0800

This systematic review examines the emerging role of Artificial Intelligence (AI) in planning and optimizing treatment for Non-Small Cell Lung Cancer (NSCLC). Focusing on patient-tailored therapy planning and enhancing treatment efficacy through advanced deep learning algorithms, we meticulously selected and analyzed thirteen high-quality research studies demonstrating AI’s integration in NSCLC management. These studies show the ability of AI to process complex clinical, radiomic, and genomic data to provide personalized therapy plans. AI technologies, such as deep learning models and machine learning, have shown exceptional promise in predicting immune responses to initial treatments, potentially revolutionizing the management of NSCLC. This review highlights AI’s transformative impact on predicting treatment outcomes, optimizing therapy regimens, and improving decision-making processes in NSCLC treatment. The collective findings from these studies reveal a significant trend towards personalized medical approaches, showcasing AI’s remarkable capacity to handle extensive datasets and forecast individual patient reactions. This reassures us about the efficiency of AI in managing complex information, thereby increasing treatment efficacy and improving patient health outcomes. However, this review also underscores the pressing need for further research and development in AI applications, highlighting the urgency and importance of this field. Integrating AI into NSCLC treatment marks a new era of precision cancer care, paving the way for more accurate, efficient, and patient-centered care. The challenges and limitations identified in this review serve as a call to action, urging the oncology community to continue pushing the boundaries of AI in cancer care. This review aims to identify the most advanced and effective technologies, enabling oncology researchers and healthcare professionals to utilize these tools without having to search through various available sources. This approach aims to streamline access to crucial information, allowing practitioners to focus on recent advancements. For this reason, the study concentrates on the last two years, which have been marked by significant integration of AI into precision medicine.

Forecasting Nonstationary Time Series Based on Dicrete Hilbert Transform

Wahyuni Ekasasmita, Khaera Tunnisa, Muh. Tri Aditya — Sun, 04 Aug 2024 00:00:00 +0800

Various predictive methods have been applied to predict the value of stocks. The purpose of this research is to implement the discrete Hilbert transform in stock returns. The ability to predict stock price movements has big implications for investors. Traditional methods are often limited in capturing the complexity of market dynamics. It was found that the proposed method obtained an average of MAE, RMSE and MAPE values of 0.02055, 0.02237, and 0.012985 which is lower than the conventional LSTM method. This research provides a new understanding of the application of discrete Hilbert transform in a dynamic global financial context.

Big Data in the Revolution of Medical Data: A Review

Amal Azeroual, Benayad Nsiri, Rachid Oulad Haj Thami, Taoufiq Belhoussine Drissi — Mon, 05 Aug 2024 00:00:00 +0800

Big Data plays a crucial role in the medical sector, fundamentally transforming the collection, organization, and interpretation of medical data. This shift significantly enhances healthcare quality, propels medical research, and improves healthcare system effectiveness. Medical Big Data comprises a vast and diverse array of health-related information, generated at an unprecedented scale and speed, including electronic health records, medical imaging, genomic data, clinical trials, and data from wearable devices. Analyzing this data can reveal vital insights into disease patterns, treatment effectiveness, and population health trends, thereby aiding in the creation of personalized medicine, predictive analysis, and innovative healthcare solutions. Effective utilization of Medical Big Data requires advanced computational and analytical methods to extract meaningful insights, thereby fueling progress in healthcare and medical research. This review aims to provide specialists with a comprehensive overview of Big Data's application in diagnostic and medical domains, including its current usage in healthcare. We particularly focus on how the integration of Big Data with artificial intelligence has led to more accurate predictive models for disease outbreaks and patient health risks, enhancing preventive care strategies. Furthermore, our analysis indicates that Big Data-driven personalization of treatment has significantly improved adherence to therapies and health outcomes in chronic disease management.

Enhancing Cold-Start Recommendations with Innovative Co-SVD: A Sparsity Reduction Approach

Manal Loukili, Fayçal Messaoudi — Fri, 16 Aug 2024 00:00:00 +0800

This research introduces a novel methodology to enhance recommendation systems, specifically targeting the challenging cold-start problem. By creatively combining Collaborative Singular Value Decomposition (Co-SVD) with an innovative sparsity reduction approach, our study significantly improves recommendation accuracy and mitigates the challenges posed by sparse user-item interaction matrices. We conduct a comprehensive set of experiments, leveraging a sample e-commerce dataset, to demonstrate the efficacy of our approach. The results illustrate the superiority of our Enhanced Co-SVD model over traditional Co-SVD, content-based filtering, and random recommendation in various evaluation metrics. In particular, our methodology excels in cold-start scenarios, providing accurate recommendations for users with limited interaction history. The implications of our research extend to practical applications in e-marketing, user engagement, and personalized marketing strategies, highlighting the potential for enhanced customer satisfaction and business success. This work represents a critical step forward in the evolution of recommendation systems and underscores the importance of addressing the cold-start problem in modern online services.

A Prevalent Model-based on Machine Learning for Identifying DRDoS Attacks through Features Optimization Technique

Sun, 25 Aug 2024 00:00:00 +0800

Growing apprehension among internet users regarding cyber-security threats, particularly Distributed Reflective Denial of Service (DRDoS) attacks, underscores a pressing issue. Despite considerable research endeavors, the efficacy of detecting DRDoS attacks remains unsatisfactory. This deficiency calls for the development of pioneering solutions to enhance detection capabilities and fortify cyber defenses against this sophisticated subtype of Distributed Denial of Service (DDoS) attacks. This study addresses this challenge by utilizing four distinct machine learning algorithms: SVM, DT, RF, and LR, supplemented by PCA. Leveraging the CIC Bell DNS 2021 dataset, our experiments produce compelling results. Specifically, both DT and RF algorithms exhibit exceptional performance with 100% accuracy and perfect F1 scores. This remarkable performance holds true with or without PCA-based feature reduction, except for dataset 4. Consequently, our research highlights the potential of machine learning in detecting and mitigating DRDoS attacks, offering valuable insights for bolstering cybersecurity measures against evolving threats.

Enhancing Mammography Models: The Impact of Radiologist Recommendations on Algorithmic Precision

Youssef Lahdoudi, Abdelghani Ghazdali, Hamza Khalfi, Nidal Lamghari — Mon, 12 Aug 2024 00:00:00 +0800

This study highlights the benefits of advanced image classification in breast cancer diagnosis and treatment. We utilize deep learning algorithms like YOLOv5 for image segmentation and Densenet121 for feature extraction from segmented regions. Our dataset includes 54,706 mammography images for comprehensive analysis. We evaluate 100 challenging cases, ensuring a balanced representation of benign and malignant instances. Validation involves 50 consensus cases. To address the class imbalance, we employ Upsampling/Downsampling. We fine-tune 14 algorithms and compare outcomes with and without radiologists' recommendations. Results show a 99.8\% AUC during testing and 59.5\% during validation without radiologists' input, which improves to 99.9\% and 93.5\% respectively with their insights. Expert guidance significantly enhances diagnostic accuracy. The study explores the interplay between algorithmic precision, dataset characteristics, and expert recommendations in breast cancer diagnosis. It provides valuable insights for leveraging technology and expert knowledge for improved medical outcomes.

Speed Control of a PMSM drive system using a nonsingular terminal sliding mode controller

Yahia MAZZI, Hicham Ben Sassi, Fatima Errahimi, Najia Es-Sbai — Sun, 04 Aug 2024 00:00:00 +0800

Due to its dependability, high accuracy, and performance, the permanent magnet synchronous motor (PMSM) is becoming an attractive option for electric vehicle traction systems. In this context, the objective is to achieve high power conversion efficiency and high mechanical speeds with great precision. Therefore, motor control is of paramount importance in EVs as a vehicle on the road is prone to various disturbances and load variations. Hence, a robust speed controller is necessary to ensure high operational performance, precise speed tracking, minimal overshoot, and disturbance rejection. In this study, a nonsingular terminal sliding mode controller (NTSMC) is proposed for the speed control of a PMSM powered by a three-phase voltage source inverter (VSI). NTSMC is a well-established method that provides high-performance control and can effectively handle parameter uncertainties and disturbances, making it highly suitable for PMSM speed control. The stability of the NTSMC is validated using Lyapunov stability theory. Finally, several simulations are performed. The proposed method demonstrates through simulations that it surpasses the conventional proportional-integral (PI) controller. Additionally, it provides precise speed tracking, high-performance control, and reduced overshoot, proving its feasibility and effectiveness.