United States Road Accident Prediction using Machine Learning Algorithms
Dominic Parosh Yamarthi, Haripriya Raman, Shamsad Parvin.PhD · 2025
Read United States Road Accident Prediction using Machine Learning Algorithms by Dominic Parosh Yamarthi, Haripriya Raman, Shamsad Parvin.PhD (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-06246.txt.
Introduction
Dominic Parosh Yamarthi, Haripriya Raman, and Shamsad Parvin predict road accidents across the United States with machine learning, in a 2025 paper. The text carries the sharper title naming the random forest method; the catalog title is broader.
Their data set covers 49 states and merges transport, police, and sensor records. Weather, human behavior, and road layout all enter the models. Regression and time series methods, led by the random forest, estimate how many accidents to expect under given conditions. Maps and seasonal breakdowns flag high-risk places and times. The authors aim the forecasts at officials who must place patrols, crews, and safety money where they do the most good.
Catalog record
| CREATOR | Dominic Parosh Yamarthi, Haripriya Raman, Shamsad Parvin.PhD |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 18 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-06246
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
United States Road Accident Prediction using Machine Learning Algorithms Dominic Parosh Yamarthi, Haripriya Raman, Shamsad Parvin.PhD Abstract Road accidents significantly threaten public safety and require in-depth analysis for effective prevention and mitigation strategies. This paper focuses on predicting accidents through the examination of a comprehensive traffic dataset covering 49 states in the United States. The dataset integrates information from diverse sources, including transportation departments, law enforcement, and traffic sensors. This paper specifically emphasizes predicting the number of accidents, utilizing advanced machine learning models such as regression analysis and time series analysis. The inclusion of various factors, ranging from environmental conditions to human behavior and infrastructure, ensures a holistic understanding of the dynamics influencing road safety. Temporal and spatial analysis further allows for the identification of trends, seasonal variations, and high-risk areas. The implications of this research extend to proactive decision-making for policymakers and transportation authorities. By providing accurate predictions and quantifiable insights into expected accident rates under different conditions, the paper aims to empower authorities to allocate resources efficiently and implement targeted interventions. The goal is to contribute to the development of informed policies and interventions that enhance road safety, creating a safer environment for all road users. Keywords: Machine Learning, Random Forest, Accident Prediction, AutoML, LSTM. I Introduction The advent of motorized vehicular transportation has dramatically transformed global mobility, shaping economies, urban landscapes, and social dynamics. However, this rapid expansion of vehicle usage has also brought significant challenges, particularly in terms of road safety. The increase in the number of vehicles on the road has led to a corresponding rise in traffic accidents, resulting in substantial human, economic, and societal costs. As a result, predicting and preventing road accidents has become a critical area of research. In recent years, the application of machine learning techniques for predicting and analyzing road traffic accidents has gained significant attention. Numerous studies have explored various approaches to enhance road safety and predict accident occurrences. The dataset used in this research was sourced from an online forum that records nationwide road accident incidents, including environmental factors and severity, as detailed in [7]. While the original dataset authors focused on classifying accident severity, our research repurposed the data to predict accident occurrences. The study in [1] analyzed and visualized the dataset, comparing classification models for accident severity. In contrast, our research conducts predictive analysis to forecast accident numbers at specific locations and times across the United States, moving beyond severity classification. This shift allows for a proactive approach to road safety, offering insights for targeted interventions and resource allocation. Our research builds on [2], where advanced transformer models predicted vehicle trajectories and accident likelihood. Unlike this vehicle-centric approach, we focused on environmental factors like weather, road types, and traffic patterns to improve prediction accuracy. By concentrating on these external factors, our model provides a broader understanding of the conditions leading to accidents, offering comprehensive predictive insights. This research aligns with broader efforts in machine learning for road safety, as seen in studies like [3], which introduced a fuzzy-RBF-based approach for accident prediction, and [4], which utilized deep learning and ensemble methods for fatality prediction. Our approach differs by focusing on predicting accident numbers rather than just categorizing severity, filling a critical gap in existing literature and offering practical applications for policy-making. Additionally, insights from this study could inform strategies similar to those in [5], where road safety prediction strategies were reviewed for IoV-enabled systems, showcasing the versatility of predictive analytics. Our methodology complements studies like [6], which analyzed machine learning algorithms for accident prediction, and [8], which used conflict coefficients for traffic accident prediction. Our research also aligns with techniques explored in [9] and [10], which determined traffic accidents using spatial-temporal techniques and aimed at reducing accident rates, respectively. Finally, our predictive approach is informed by methodologies in [11] and [12], which applied 3D scene reconstruction and analyzed environmental factors in traffic accidents. By integrating these varied methodologies, our research offers a comprehensive perspective on accident prediction. The use of data visualization and machine learning, as seen in [13] and [14], and the comparative analysis of algorithms in [15], further underscores the importance of our predictive model, which aims not only to classify but also to forecast accident occurrences, enabling more effective road safety interventions. Our work stands out by focusing on predicting the actual number of accidents at specific locations and times across the United States rather than just classifying accident severity. We integrate environmental factors using Random Forest regression to enhance prediction accuracy. This shift from classification to prediction allows for more effective resource allocation and targeted interventions, making our approach more practical and impact in improving road safety compared to prior research. We utilized a Random Forest Regressor to predict the number of traffic accidents at specific locations and times across the United States. This machine learning technique was chosen for its ability to handle large datasets and its robustness in capturing complex patterns within the data. By leveraging the ensemble learning approach of Random Forest, we achieved high accuracy in forecasting accident occurrences, offering valuable insights for proactive road safety interventions. II Related Work With the help of data science and machine learning models, the key achievements like different kinds of analysis, identifying the reason for accidents were achieved by the use of historical data. This section delves with the most latest works which is used to achieve the goal providing an overview of significant studies within the research community [1]. Sobhan Moosavi and team proposed the DAP model [7]. The maximum F1 score achieved by the model involving several neural network layers. While promising, the authors recommend to incorporate the latest methodologies for data cleaning and prepping steps which would significantly improve the accuracy. [7]. Honglei Ren and colleagues introduced a technique called LSTM, which models long-term dependencies in sequential data, for assessing hazards in traffic accidents. Their study, showing the prediction accuracy(RMSE) of 0.034, revealed the non-uniform distribution of traffic accidents in space and time.[8]. However, limitations were noted in the dataset scope, highlighting specific restrictions and temporal analysis. [9]. Lu Wenqi designed the TAP-CNN model, utilizing a Convolution Neural Network (CNN) architecture, achieving a prediction accuracy of 78.5% [10]. The evaluation employed United States I-15 highway accident data within a restricted geographic area. Notably, the study focused on a small region, potentially limiting the generalizability of the model [11]. Senk and his team explored a model that predicts accidents by detecting hazardous areas through spatial pattern analysis of historical accident data [12]. But the model did not consider key factors such as time and weather, which could be major causes of accidents [13]. Tessa K. Anderson proposed k-means algorithm for hotspot profiling of accidents, based on data from the UK Police Department. Their research showed accident hotspots and patterns by mapping their density in various regions. The consideration of whether to develop region-specific or generalized systems for accident prediction is discussed [14, 15]. III Methodology III- 1 Dataset This research utilizes the dataset of accidents occurring in the United States, which is continuously updated with records obtained from the MapQuest and Bing APIs. The dataset snapshot used for this study covers the period from February 2016 to March