Telegram Facebook

EACE Journal — Vol. 15, No. 1

Identification of Accident Black Spot Locations Using Statistical Modeling: A Case Study of Addis Ababa

Mitiku Dinsamo

Abstract

Abstract The research aimed to identify accident blackspot locations within Addis Ababa’s Road network and develop an accident prediction model using statistical modeling techniques. Out of fifty-two road segments analyzed, twenty were identified as black spots, characterized by high incidences of speeding vehicles relative to the proposed speed limits. Specifically, nine black spots were found on the road from Bole to Meskel Square, seven from Mexico to German Square, and four from Sarbet to Sost Kutir Mazoria. The study utilized various explanatory variables to develop the accident prediction model, including adjusted average peak hour volume, 85th percentile speed, availability of U-turns, and the number of access points. The model was derived using a generalized linear model (GLM), specifically the negative binomial (NB) regression analysis, which was deemed appropriate due to over dispersion in the data. Key findings from the statistical analysis revealed that the 85th percentile speed had a significant positive relationship with accident frequency, with a coefficient of 0.105 and a p-value of 0.000, indicating high statistical significance. This means that for each unit increase in the 85th percentile speed, the expected log count of accident frequency increases by 0.105. Additionally, the availability of U-turns was found to increase accident frequency, while the number of access points had a negative effect, with a coefficient of -0.158, suggesting that more access points lead to fewer accidents. To evaluate the model's goodness of fit, several measures were employed, including Pearson chi-square, deviance, Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC). The results indicated that the negative binomial model fit the data well, with Pearson chi-square and deviance values close to the degrees of freedom, confirming the model's adequacy. Furthermore, comparative analysis with the Poisson model showed that the negative binomial model provided a better fit, as evidenced by lower AIC and BIC values. The findings underscore the importance of considering high-risk factors like 85th percentile speed and the availability of U-turns in traffic management and road safety strategies. The developed accident prediction model offers a robust tool for identifying and mitigating accident blackspots, ultimately contributing to safer road networks in urban areas. Keywords: Accident Black Spots, Statistical Modeling, Traffic volume, Spot Speed, Road Safety.

Read Full Article (PDF)