Comprehensive approach to predictive analysis and anomaly detection for road crash fatalities

AIP Advances, Год журнала: 2025, Номер 15(1)

Опубликована: Янв. 1, 2025

Since traffic accidents are a major global cause of injury and death, it is essential to comprehend reduce their effects. Finding high-risk areas creating focused interventions increase road safety made possible by the research’s analysis numerous variables that affect number fatalities in crashes, including weather, features, geographic locations. To further contribute overall objective building safer transportation networks for everyone, application predictive models anomaly detection techniques enables proactive steps avert collisions lower on our roadways. With main improving safety, thorough approach was put into place evaluate data from forecast deaths, identify abnormalities. Using multimodal method, research first combines two datasets based coordinates: crash count data. This integration makes easier grasp various aspects comprehensively. These factors include regions. A Random Forest Regression model trained estimate deaths arising crashes after preprocessing, which includes feature selection encoding. The accuracy power assessed through utilization Mean Squared Error measure. determine most important impacting importance also carried out. find anomalies or outliers take preventative action impact accidents, utilizing an Isolation utilized. Through possibility highlighting regions with increased risk problems quality, this part improves comprehension unexpected events accident For comparison analysis, other such as Auto Regressive Integrated Moving Average Support Vector used addition model. root mean squared error statistic analyze these models’ performance applicability real-world scenarios. They provide different viewpoints prediction mortality accidents. study’s findings highlight significance using data-driven strategies successfully solve issues related safety. offers policymakers, authorities, advocates practical insights sophisticated machine-learning algorithms integrating multiple datasets. Road can be decreased systems established have been created tool identifying allocating resources targeted improvements. enhance results, emphasizes need interdisciplinary partnerships decision making. open door evidence-based initiatives lessen effects save lives roads analytics modeling.

Язык: Английский

---

Enhanced Image Tampering Detection using Error Level Analysis and CNN

Engineering Technology & Applied Science Research, Год журнала: 2025, Номер 15(1), С. 19683 - 19689

Опубликована: Фев. 2, 2025

This paper introduces a novel approach to image tampering detection by integrating Error Level Analysis (ELA) with Convolutional Neural Network (CNN). Traditional forensic methods, such as ELA and Residual Pixel (RPA), often struggle detect subtle or advanced manipulations in digital images. To address these limitations, this method leverages highlight compression-induced variations CNN extract classify spatial features indicative of tampering. The dataset, consisting both authentic tampered images, was preprocessed generate representations, which were then used train model designed distinguish between manipulated regions. Extensive experimentation performed on the CASIA v2.0 demonstrating significant improvements accuracy, precision, recall. proposed framework achieved accuracy 96.21%, outperforming established deep learning models VGG16, VGG19, ResNet101. These results underscore potential combining advancing forensics, offering robust solution ensure integrity content an era sophisticated manipulation.

Язык: Английский

---

A Novel Non-Iterative Deep Convolutional Neural Network with Kernelized Classification for Robust Face Recognition

Engineering Technology & Applied Science Research, Год журнала: 2024, Номер 14(5), С. 16460 - 16465

Опубликована: Окт. 9, 2024

Deep Convolutional Neural Networks (DCNNs) are very useful for image-based pattern classification problems because of their efficient feature extraction capabilities. Although DCNNs have good generalization performance, applicability is limited due to slow learning speed, as they based on iterative weight-update algorithms. This study presents a new noniterative DCNN that can be trained in real-time. The fundamental block the proposed fixed real number-based filters convolution operations multi-feature extraction. After finite number layers, nonlinear kernel mapping along with pseudo-inverse used extracted vectors. DCNN, named Kernelized Classification (DCKC), noniterative, mask coefficients its numbers. function predefined parameters DCKC does features, and find output weights. was evaluated benchmark face recognition databases, achieving better results establishing superiority.

Язык: Английский

---

Autofocus Vision System Enhancement for UAVs via Autoencoder Generative Algorithm

Engineering Technology & Applied Science Research, Год журнала: 2024, Номер 14(6), С. 18867 - 18872

Опубликована: Дек. 2, 2024

The Autofocus (AF) technology has become well-known over the past four decades. When attached to a camera, it eliminates need manually focus by giving viewer perfectly focused image in matter of seconds. Modern AF systems are needed achieve high-resolution images with optimal focus, and very important for many fields, possessing advantages such as high efficiency autonomously interacting Fenvironmental conditions. proposed vision system Unmanned Aerial Vehicle (UAV) navigation uses an autoencoder technique extract features from images. system's function is monitor control camera mounted drone. On dataset, model exhibited amazing 95% F-measure 90% accuracy, so can be considered robust option achieving precision clarity varying conditions since effectively identify features.

Язык: Английский

---

Development of a MEMS-based Piezoresistive Cantilever Sensor for Lead (Pb(II)) Detection in Drinking Water

Engineering Technology & Applied Science Research, Год журнала: 2024, Номер 14(5), С. 17330 - 17336

Опубликована: Окт. 9, 2024

One of the most hazardous pollutants natural water resources is lead -Pb (II)- which poses a significant threat to human health and environmental safety. The accumulation this heavy metal in an organism affects number systems particularly dangerous for children. At low levels intake over short periods, it induces diarrhea, abdominal pain, renal damage, with potential fatal outcomes extreme cases. principal sources pollution are industries, coal-fired power plants motor vehicles. In response critical demand effective detection, researchers have developed advanced Micro-Electromechanical Systems (MEMS) piezoresistive cantilever sensors that make use chelating properties Ethylenediaminetetraacetic Acid (EDTA) superior electrical reduced Graphene Oxide (rGO). It has been proven composite can be effectively immobilized on MEMS surface, enabling selective removal Pb (II) ions from wastewater. This adsorption process exerts stress surface cantilever, resulting variations resistance subsequently measured. A sensitive sensor developed, offering as monitoring tool samples. demonstrated high sensitivity selectivity, detection limit 1 ppb linear range 10-100 ppb. novel approach significantly enhance provide substantial benefits public by real-time, on-site mapping contamination across aqueous environments. technological advancement surveillance domain offers new perspective safety reduction hazards associated consumption.

Язык: Английский

---