Volume: 19 Issue-01 (January-June) 2024

---

PREDICTION OF ROAD ACCIDENTS USING MACHINE AND DEEP LEARNING TECHNIQUES

ABSTRACT

Globally, the occurrence of traffic collisions has shown a noticeable increment within recent years. It has become a major issue due to the vast amount of auto accidents that happen annually. It is horrifying and completely unbearable for its citizens to lose their lives in auto accidents. This paper aims to examine traffic collisions that occur at the federal, state, and local levels in India. By employing machine learning and deep learning methodologies, this research conducted a comprehensive investigation into traffic accidents and determined the precise accident sites. Combining deep learning and machine learning technology for intelligent accident detection is advised as part of an all-encompassing approach to enhancing traffic safety. In order to create a worldwide safety performance function (SPF) that can project collision rates for diverse roads in different places, the research investigates the idea of applying a machine learning technique. The study investigates the synergy of ML and DL in predicting and analyzing accident locations using a broad dataset supplied from Kaggle, which includes variables ranging from geospatial coordinates to weather conditions. This study focuses on forecasting automobile accidents based on road type, including asphalt, gravel, and earth roads. The study includes machine learning approaches such as data preprocessing and model training, as well as historical accident data, road infrastructure characteristics, and environmental variables. Given the increasing number of vehicular collisions and casualties in India, the issue of road accidents is a matter of utmost importance. India faces a number of challenges in ensuring road safety. Rapid urbanization, increased vehicle density, diverse road users, and insufficient infrastructure all contribute to a high accident rate. Understanding these limitations is crucial for developing effective treatments. The age group between 30 and 59 years old holds the highest susceptibility, wherein males encounter a larger percentage of both fatalities and injuries. Traffic accidents are more prevalent in bad weather and during business hours. The research technique entails gathering and combining several statistics, such as accident severity, number of victims, latitude, longitude, vehicle count, and vehicle characteristics. In this case, it will be good to investigate the frequency of accidents so that we may utilize this information to build ways to reduce them. RESNET, Random Forest, SVM, and LSTM are therefore used to increase the precision and effectiveness of event detection.

FULL TEXT

PDF