Scholarlink

<p>One of the main challenges in traffic management in a smart city is traffic congestion, which has several major causes. Numerous causes, including inclement weather, excessive speed, road friction, and busy hours, can lead to unexpected accidents that result in injuries, impairments, and occasionally fatalities. As such, it becomes imperative to ascertain the correlation between features or pinpoint the element that might potentially cause an accident in a certain scenario. A few techniques are available in the existing&#xa0;system to verify the hotspot rating. However, when using local spatial indices to find hot areas, false discovery rates are typically greater. In this work firstly, A geographic information system(GIS)-based Gaussian probability density estimation (GPDE) algorithm was used to identify hotspots. Second, Moran's K-mean clustering (MK-mean) was used to determine the statistical importance of the hotspot clusters. In the end, the Gradient Boosting Convolutional Neural Network (GBCNN) classifier identified hotspots. This classifier is quite accurate when determining danger patterns in an area, such as hot parts, usual locations, and cold places. Other performance evaluations confirm that our suggested technique performs better than any other strategy currently in use, as demonstrated by the experimental findings.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A novel analytical framework to identify and classify accident hotspots integrating gradient classifier and spatial clustering

  • 期刊论文

  • 发表日期：

    2025年1月13日

查看全文
  • Nidhi Singh,
  • Manoj Kumar

摘要

One of the main challenges in traffic management in a smart city is traffic congestion, which has several major causes. Numerous causes, including inclement weather, excessive speed, road friction, and busy hours, can lead to unexpected accidents that result in injuries, impairments, and occasionally fatalities. As such, it becomes imperative to ascertain the correlation between features or pinpoint the element that might potentially cause an accident in a certain scenario. A few techniques are available in the existing system to verify the hotspot rating. However, when using local spatial indices to find hot areas, false discovery rates are typically greater. In this work firstly, A geographic information system(GIS)-based Gaussian probability density estimation (GPDE) algorithm was used to identify hotspots. Second, Moran's K-mean clustering (MK-mean) was used to determine the statistical importance of the hotspot clusters. In the end, the Gradient Boosting Convolutional Neural Network (GBCNN) classifier identified hotspots. This classifier is quite accurate when determining danger patterns in an area, such as hot parts, usual locations, and cold places. Other performance evaluations confirm that our suggested technique performs better than any other strategy currently in use, as demonstrated by the experimental findings.