Vehicle Road Accident Prediction Model along Federal Road FT050 Kluang-A/Hitam-B/Pahat Route Using Excess Zero Data

Abstract

Traffic accidents have become a major socio-economic problem in Malaysia as it is the primary cause of mortality. Over 60 percent of these fatal accidents occurred on rural roads. Nearly half of all fatalities took place on federal roads and over a quarter happened on state roads. It is also estimated that about 2 percent of the country’s Gross Domestic Product (GDP), or approximately RM 9 billion, is lost through road accidents. Previous studies managed to develop several models for modelling the occurrence of accidents, but most of these models have plenty of deficiencies. The following study focuses on stochastic regression models, such as Poisson, Negative Binomial, Zero-Inflated Poisson and Zero-Inflated Negative Binomial with excess zero outcomes on the response variables. Furthermore, in order to specify the regression relationship with a sophisticated result, R-statistical programming is used. The method used is also the updating approach in predicting potential road accidents, which can also produce an accuracy probability of hazardous locations. Based on road accident data collected over a five-year period from 2010 to 2014 at Federal Road F0050: Kluang-A/Hitam- B/Pahat in Johor, Malaysia, results of this study show that Zero Inflated model performed better, in terms of the comparative criteria based on the AIC value.

Modelling Road Accident Fatalities with Underdispersion and Zero-inflated Counts

“…For example, the Poisson, NB, ZIP and ZINB were used to model of numbers of RAFs on the highest accident road F0050 in Malaysia [25] and in the Oromia region, Ethiopia [19]. Modeling zero inflated regression of Poisson and NB of the number of RTFs on the road F001 - Jalan Jb Air Hitam and on the road FT050 - Kluang-A/Hitam-B/Pahat, Malaysia was presented by [24] and [26], respectively. The results suggest that the zero inflated version models offered better statistical performance than the traditional models.…”

“…Differences in the distribution of accidents, deaths, and injuries could be attributed to the variation between roads and their surroundings. Several studies estimate the frequency of RTAs and/or RAFs by traditional Poisson regression [17], Negative Binomial (NB) regression [18][19][20], and Conway-Maxwell Poisson (CMP) regression [21][22][23], as well as estimate the count data with excessive zeroes by zero-inflated regression models [24][25][26]. However, predictive models for deaths and/or injuries are limited in Thailand road safety.…”

Regional Perspective of Safety Performance Functions and Their Application to Florida Intersections in Suburban Residential and Urban General Context Classification Categories

“…ZIP and ZINB models can be effective at modeling data with excessive zeros since they allow for different sets of variables to model the zero state and the count state ( 14 ). The equations for the ZIP and ZINB models are the same as the equations for the Poisson model (Equation 2) and NB model (Equation 3), respectively, with the zero-inflated models having an additional model of the same form for the excess zero counts.…”

“…To address this issue, ZIP and ZINB models can be used for SPF development. These models allow for different sets of variables to model the zero state and the count state ( 14 ). A study conducted in Malaysia showed that a ZINB model outperformed Poisson and NB models as indicated by a lower Akaike information criterion (AIC) value ( 14 ).…”

Analysis of Severity Level Types and Trends in Road Accident Cases at Johor Inter State Road Using Analythical Hierarchy Process (AHP) And Geographical Information System

“…The Average Annual Daily Traffic (AADT) and speed of vehicles were used in the research for the traffic information [5]. In year of 2010, 2011, 2012 and 2013, the AADT was provide by the ministry of works Malaysia Highway planning unit road traffic volume Malaysia [6]. In this organization, it conducts a bi-monthly study at each year in a specific station in April for the first half data and in October for the other half year data [4].…”