#### 本站介绍

360论文服务中心是最受欢迎的论文发表与论文编辑服务网站。

360论文服务中心于2000年创建，注册用户量已突破152万人，并帮助近380万人次顺利发表论文。14年来，360论文服务中心始终遵循热情快捷、安全可靠的服务宗旨，深受广大网民青睐。360论文服务中心主要设有论文发表部、论文创作部、期刊合作部、论文采编部、技术运营部和市场推广部等多个部门，是目前国内论文行业，规模最大、服务人员最多的正规网站。

#### 创作发表说明

1. 如果您没有论文，还要评定职称，需要发表论文，请联系我们，客服人员会及时处理；

2. 教授、博士组成的论文指导团队，专业打造高品质论文；

3. 合作期刊，全国最全，与杂志社关系稳定，保证刊期。

目录

1 Introduction 1

2 Literature Review 3

2.1 The Theoretical Foundation 3

2.1.1 Bernoulli Trial 3

2.1.2 Negative Binomial Distribution 3

2.1.3 Poisson Distribution 4

2.1.4 Zero-Inflated Poisson Distribution 5

2.2 The Precious Research about Car Accident 7

2.2.1 Accidents related theories 7

2.2.2 Study Review 10

3 Methodology 11

4 Data description 12

4.1 Data source 12

4.2 Variables 12

5 Data analysis 13

5.1 For body injury 13

5.1.1 The regression of Binomial family 13

5.1.2 The regression of Poisson family 13

5.1.3 The regression of zero-inflated Poisson model 13

5.2 For accident 13

5.2.1 The regression of Binomial family 13

5.2.2 The regression of Poisson family 13

5.2.3 The regression of zero-inflated Poisson model 13

6 Conclusion and suggestion 14

Reference 15

1 Introduction

Road traffic accidents, refers to the vehicle drivers, pedestrians, passengers and other road transport-related activities on the officers for violation of traffic regulations, rules of behavior, negligence, personal injury or property damage caused by an accident. Since the invention of the automobile since the number of deaths in road accidents around the world more than three million people, the number of deaths over the same period more than war. Number of traffic fatalities attributable to non-natural death of about 1/4, has become the world's largest pollution, traffic accidents to society, the family is enormous harm and long-term.

Go to mention the level of urbanization, traffic demand; urban expansion, road construction increased, denser road network; development of automobile industry, the automotive market strong increase in vehicle ownership. Road traffic is its easy and fast characteristics in the social and economic life is playing an increasingly important role. But traffic safety issue has become a major social issue in today's world. Safe, efficient and economical transportation service object is made on the basic requirements for the transport sector, but also the pursuit of goals transport workers. How to effectively solve the problem of traffic safety is an urgent need to resolve the issue, attracted world transportation agencies and researchers.

2011, the United Nations World Health Organization issued a second report on the global status report on road safety. Reported that there are nearly 130 million people each year are killed in road traffic accidents, which killed three thousand people every day. Suffer non-fatal injuries due to traffic accidents have twenty million to fifty million people, these injuries are an important factor in causing global disability, of which more than 90% deaths occur in low and middle income countries. If you do not take immediate action, as the global improvement in the level of motorization, the number of road traffic casualties will continue to rise. If current trends are not curbed, it is expected that by 2030, road traffic injuries will be the ninth leading cause of death from the present up to fifth leading cause of death, killing nearly 240 million deaths annually. The main reason is the rapid increase in global motorization, while at the same time the corresponding road safety strategy and planning did not improve traffic safety. March 2010, the UN General Assembly adopted a resolution A/RES/64/255 announced 2011-2020 as the Decade of Action for Road Safety. The goal is at the national, regional and global levels within the context of more activities are expected to stabilize and further reduce global road traffic deaths.

With the economic and social development, automotive technology has brought high accident rate forcing countries invest a lot of manpower and material resources to conduct a traffic safety study, in which the accident analysis and prevention of particular attention has been paid, because of traffic accidents not only cause economic, and property loss, and related to people's lives. Accidents related to many factors, factors such as road, traffic facilities, traffic volume and traffic composition, environment and climate, the driver's driving behavior. By studying the status and characteristics of road traffic accidents, road traffic accidents found to affect the plains of the main factors, using statistical regression method highway accident prediction model. By accident prediction can grasp the future road traffic safety situation and not road safety hazards, so that timely, targeted to take appropriate measures, through the effective control of a variety of factors, to reduce traffic accidents, improve road safety purposes.

Road traffic accidents will increase as economic development, the increase in car ownership, there is a growing trend. In road traffic planning, design, management, regulatory and education, road safety has become increasingly important scientific decision-making. Therefore, good highway traffic safety analysis and research work to improve the safety of road infrastructure and improving traffic safety management level has very important significance. It is necessary to sort out and road safety research, the establishment of a scientific and effective evaluation, forecasting, analytical methods, and thereby reducing traffic accidents. Road traffic accident research contribute to a comprehensive, fully reveal the accident in space, time, and other distribution, analysis of accident-prone locations causation. By road accident identification, investigation and causation analysis, posed to targeted improvement measures can significantly reduce less investment for the entire road accident rate effect; you can achieve greater economic and social benefits.

2 Literature Review

2.1 The Theoretical Foundation

Count data widely exists in the accident analysis, finance, insurance, social sciences and biomedical research fields. Currently, count data statistical research has become a hot topic issue. Traditional and fitting count data are commonly distributed Poisson, binomial, negative binomial, generalized Poisson distribution, etc., but in the actual problem count data often exhibit excessive zero observations case, If you are still using these types of model analysis will lead to biased results of statistical inference. It is better to solve such problems, zero-inflated regression models have been proposed and used to analyze and research data containing excessive zero issue. Road traffic accidents with discrete, stochastic and independence, can occur at any one space unit, the number of accidents within the cell count data model can be used to describe common data model used in traffic counts have Poisson, negative binomial and zero-inflated Poisson model.

1992 Lamber defines a zero-inflated Poisson regression model, this model gives a parameter estimation method, standard errors and confidence intervals, also on the wiring board soldering defects an illustrative example. 1994 Miaou this model will be applied to the risk analysis, he used the Poisson regression model, the negative binomial regression model and the zero-inflated Poisson regression model truck accident on the relationship between risk and road conditions were analyzed and compared fitting results of three models, the sensitivity analysis of risk factors that the three models have their own strengths: Poisson regression model should be considered first, if the data exhibit heterogeneity you should consider the negative binomial regression model and zero-inflated Poisson regression model. 2005 Karen CH Yip, KelvinK W yau detail in the article compares the number of insurance claims sample Poisson distribution, negative binomial, zero-inflated Poisson, zero-inflated negative binomial, zero-inflated Poisson and zero mixing expansion generalized Poisson distribution, and their respective corresponding regression model fitting results, showing the application of the model number of zero expansion.

2.1.1 Bernoulli Trial

Under the same conditions the experiment repeated n times is called Bernoulli trial. "Under the same conditions" is equivalent at the respective test results are not affected by other experimental results. Bernoulli trials to determine whether the key is the probability of the event A for each test constant, and the test results with each other irrespective of the results of tests, is repeated for a series of experimental tests, is not a test, and many times, but pay attention to the probability of occurrence of a repeating event has no effect on each other.

2.1.2 Negative Binomial Distribution

Negative binomial distribution suitable to describe organisms in spatial, temporal aggregation of problems, such as the distribution of snails in the soil, insect’s spatial distribution. Where is a clustered index (a measure of the distribution of the degree of dispersion). Unlike the Poisson distribution of the mean and variance equal to the variance of the negative binomial distribution is greater than the mean, a single point can be anticipated from this: it is more suitable than the Poisson distribution with a heterogeneous sample data.

Tllomas (1987), Renshaw (1994), Habermall (1996) on the negative binomial distribution applications in the accident conducted in-depth research. Using negative binomial regression parameters were calculated for traffic analysis. This method is a multivariate regression method that is the promotion of Poisson regression, but avoids the multiple linear regression and Poisson regression shortcomings. Kestemont and Paris (1985) defines a Poisson distribution with a mixture of a large class of probability distribution functions, and propose an effective method to estimate parameters. Ruohonen (1987) proposed a structure for the three-parameter gamma function is a function of the Poisson distribution, while using actual loss data and the two parameter Poisson model structure functions i.e. negative two models were compared, have been relatively satisfactory results. Panjer (1987) using the generalized poisson pascal distribution to establish the number of car accidents and claims model fitting effect is ideal. Consul (1990) generalized Poisson models were fitted, but fitting effect as Kestemont and Paris (1985) model. In addition, Elvers (1991) found that the generalized Poisson distribution cannot be a very effective way for a group of car accidents to fit the data. Islam and Consul (1992) recommended that the Consul proposed model as a probability distribution models to fit the number of car accidents and insurance claims. They were used Panjer (1987) data and Gossiaux and Lemaire (1981) to fit the data and found that without considering the case of zero-class fitting results were better. However, Sharif and Panjer (1993) found that in the presence of a series of model fitting Consul Defects such as parameter space limitation, the derivation of maximum likelihood estimation process of the emergence of some theoretical issues and so on. They are more inclined to use some simple probability models such as generalized poisson-pascal model or Poisson inverse Gaussian model to fit the data. Denuit (1997) proved Lefévre and Picard (1996) proposed poisson-goncharov distribution can be described as a pilot year, the number of claims probability model, and gives the estimation method.

Negative binomial regression model requires the mean and variance are equal. If the samples are too scattered distribution, Poisson distribution cannot meet the conditions of the accident prediction; this time requires the use of negative binomial regression model. Negative binomial distribution is a continuous mixed Poisson distribution; Poisson regression model is limited in that the mean and variance equal to the requirement, when the accident too discrete sample distribution, variance is much larger than the mean for the Poisson distribution is no longer satisfied. In order to eliminate or reduce such variance greater than the mean of the negative impact, Haur E (1988), Hinde J (1998) proposed using negative binomial regression model to replace the Poisson regression model.

The meaning of the function: When the study area appear N random distribution point, the spatial units in points fall into one of the probabilities. The formula is the gamma function; non-negative index is called the spreading parameters; Poisson distribution as parameters characterizing each statistical unit desired number of accidents, is the variance.

2.1.3 Poisson Distribution

Poisson distribution is a common discrete probability distribution, by the French mathematician Simeon-Denis Poisson published in 1838, describes the unit of time for the number of occurrences of random events. If a service facility within a certain time to reach the number of telephone switches the number of calls received, number of guests waiting car sites, the number of machine failures, the number of natural disasters and so on.

Poisson regression model prediction is based on the principle of the Poisson distribution, Poisson distribution strength parameters for modeling disobedience against the overall distribution of the accident situation. American scholars Miaou other accidents 1985-1987 based on U.S. data, analysis and road train accident relationship between the structure proved Poisson regression model is more than a linear regression model for prediction of traffic accidents.

Poisson distribution is suitable to describe the unit of time (or space) random number of events, the probability function is:

Where, is defined as the time in the statistical spatial unit i occurred in the number of accidents on, is the Poisson distribution parameter and exponential relationship, is associated with the spatial unit vector of explanatory variables, is the coefficient corresponding to the explanatory variables .

Poisson distribution has only one parameter into both the Poisson distribution of the mean, the variance of the Poisson distribution is. Poisson distribution of additive: if and independently, then .

In theory, the Poisson distribution should have the same mean and variance, but the actual number of accidents study sample variance is often much larger than the average, the occurrence of super-spreading phenomenon. Solution to this problem is to use negative binomial model to fit the data. For the random distribution of the spatial position of the accident, the first in space R N times on randomly generated incidents, and use the process to generate random binomial distribution point mode.

Poisson regression model is built on the basis of the Poisson distribution; specialized analysis should be variable for the count variable regression model, which allows each sample has a different mean long, detailed description about this model can refer Cameron1998 year’s writings. Tllomas (1987), Renshaw (1994), Haberrnan (1996) also on the accident and insurance aspects of the research.

2.1.4 Zero-Inflated Poisson Distribution

Count data in the social sciences actual study, the observed number of events are often found contains a large number of zero values that many individuals in the observation unit of observation time, space, does not occur within the corresponding area of random events, such as a year of hospitalization, divorce number of times in jail, number of children, number of abortions, etc.; forest fires in the study, the vast majority of the year day, the corresponding region is no fire; in the non-life insurance actuarial, discrete causes of the phenomenon over Many, such as companies and policy holders increased awareness of risk prevention, risk most insurance policies do not occur, or because insurance companies use deductibles or no claims discounts and other terms of the occurrence of the insured causes many minor incidents without will claim that these phenomena will lead to an obvious result - zero expansion phenomenon; in medical research, and often encounter observed number of events contains a lot of zero, that many observers observing the individual in the corresponding period of unobserved events, such as myocardial ischemia in patients with coronary heart disease observed number of segments.

Poisson regression modeling is commonly used discrete data model is widely used in various application fields of science, but the premise of the model is the use of the conditional mean is equal to the conditional variance. But in practice, because the count data contains too much 0 observations, makes the model the conditional variance is much larger than the conditional mean, at this time there will be zero expansion phenomenon. To solve this problem, Lambert was first proposed zero-inflated model (Zero-inflated Model, ZIM) is a special case: zero-inflated Poisson model (Zero-inflated Poisson, ZIP). The basic idea is the whole count data is divided into two processes: 0 Poisson counting process and the counting process. With this mechanism, we can establish the process for these two mixed probability distribution. Negative binomial regression and Poisson regression as a generalized Poisson regression promotion, likewise, may be zero-inflated Poisson model leads to zero-inflated negative binomial model and zero-inflated generalized Poisson model.

1960s, zero expansion phenomenon caused some attention of scholars (Cohen, 1963; Johnson & Kotz, 1969). Lambert (1992) was the first time the class model into practice in the past, and manufacturing of welding defects data were fit analysis. For the treatment of this phenomenon, some scholars have proposed Hurdle model, applied economics zero expansion phenomena (Mullahy, 1986). Denuit and Marechal, etc. (2007), Winkelmann (2008), respectively, in their monograph on the zero expansion phenomena are discussed and analyzed. For the field of non-life insurance actuarial zero expansion phenomenon questions, Yip and Yau (2005) model using different zero-inflated data to a group of auto insurance claims frequency were fitted comparative analysis; Jong and Heller (2008) in the generalized linear model Life within the framework of the statistical analysis of the sort, combined with the actual data and application of zero-inflated models were analyzed. Then he made another deal with the phenomenon of the model - zero-inflated Poisson model (ZIP), used in electronics manufacturing quality control, namely the introduction of covariates on the zero-counting process and the counting process to establish a non-zero probability of mixing distribution, the establishment of a ZIP model with covariates (Lambert, 1992). For a zero expansion phenomenon, many scholars have proposed various types of statistical models. If Czado and Erhardt, etc. (2007) proposed zero-inflated generalized Poisson model to solve the death of the fetus movement and expansion of the number of zeros in the data problem; Efron (1986) have proposed zero-inflated Poisson model pairs; Yip and Yau (2005) is right auto insurance claims in the frequency data using zero-inflated Poisson (ZIP) model, zero-inflated negative binomial (ZINB) model, zero-inflated generalized Poisson model and zero-inflated Poisson models were fitted double comparative analysis. The zero-inflated Poisson model is extended to zero-inflated negative binomial model (Zero-inflated Negative Binomial, ZINB), and using BHHH method to estimate the standard error of the model parameters, applied to consumer banking bad credit records research (Greene, 1994) . ZINB model is a Poisson model and the Negative Binomial model and technological development, to make up for the negative Poisson model or two models in the analysis of zero-inflated when the lack of structural data, count data can explain the excessive value of zero, so that the dependent variable identification of the real value of zero as possible, but also to make more efficient estimation results with and without bias, to obtain reliable parameter estimation and hypothesis testing, in order to help researchers answer a series of practical significance and the traditional model cannot answer the question. Shankar (1997) that the model can be considered as unconditionally zero-inflated count a dual process model because the model of the part from the zero point zero stationary distribution, and the other part of the counting process of the zero standard. He also suggested by logistic method to distinguish between the two parts of zero, the relevant literature can be found in Lambert (1992). Lee (2002) with unconditional zero-inflated models to fit car collisions, collision risk among a larger number of accidents by drivers to describe the Poisson process. Miaou (1994) using the same method trucks fitted zero-inflated number of accidents. Most models can be unconditionally zero-inflated EM algorithm or Newton's iterative method to estimate parameters (Lambert, 1992; Heliborn, 1994; Shankar, 1997).

Because the actual problem may occur with 0 too many discrete data, if fitted with a Poisson distribution, which did not occur and the expectations of an incident with the observed frequency of the frequency difference between the larger, so you can consider using zero-inflated Poisson model fit). In the model predictions, the zero accident data is large, the Poisson regression model and the negative binomial regression model from the number of traffic accidents model cannot explain this phenomenon, leading to model error; zero-inflated negative binomial regression model using the traffic the number of accidents model will be the number of accidents. situation is set to a zero probability value, and assuming that the probability to obey normal or Logistic distribution, and can better deal with the phenomenon of excessive zero value.

Zero-inflated Poisson distribution can be viewed as a zero probability and Poisson distributions concentrated mixture. About the zero-inflated Poisson distribution model of the most famous article is Lambert (1992) published by zero-inflated poisson regression, with an application to defects in manufacturing. 1994 Heilbron using this model to assess the high-risk heterosexual behavior.

Where, Pi is called zero expansion parameter, which means that the zero count generated by the probability distribution. Obviously more than the Poisson distribution of a parameter , the parameter is used to indicate the generation of zero upon exposure to their and is not exposed in two parts, is a probability, where the exposed portion is in front of said additional zeros.

2.2 The Precious Research about Car Accident

2.2.1 Accidents related theories

1 Mechanism of traffic accidents

Traffic accidents mainly due to the people, cars, road environment consisting of dynamic systems imbalance results. People are the backbone of the whole system; road is the basis for the entire system; car is complete transport function tool. First, define the way people source of information and vehicle driving state, the human sense organs constantly ingest through road information, and after processing, and then through the motor organs to control the vehicle, the vehicle to adapt to changing road conditions and environmental conditions, while sense organs and uptake of vehicles on the road constantly adapt to the situation information, and further fine-tune the vehicle, so that the cycle reaches safe driving purposes. Person as a center of the system, vehicles and roads are to serve, therefore, people are active, conscious, is the most active factor in the system, random large; while the vehicle and the road is an objective, unconscious. Therefore, in the event of an accident, they tend to define a simple recognition that: the underlying causes of the accident is artificial, is the driver's mistakes and errors. Accidents Cause statistical results are generally as follows: man-made accidents accounted for 80% to 85%, vehicle accidents caused by factors accounting for about 5% to 10% of road accidents caused by factors accounted for about 10%.

2. Traffic safety research.

In recent years, the number of fatal road accidents are increasing. Egypt per million kilometers of mortality was significantly higher than the other three Arab countries and six G-7 countries. In the northeastern part of the Italian province of Udine, road traffic fatality rate than the full 37% higher than the Italian average. In Saudi Arabia (KSA) Riyadh, Kingdom, the annual registration of all accidents, serious incidents accounted for more than 50%. UK Department of Transport (2004) statistics show that the UK each year more than 3,000 people died in road accidents. Singapore in more than one-year period, a total of 226 people were killed, 82.3% of the victims were male, with an average age of 31 years old. Between 2005 and 2007, the region in Romania pedicle mish deaths, a total of 407 cases of traffic accidents by forensic autopsy.

3 risk of traffic accidents

In general, the head (86.7%), followed by the thoracic (67.7%) and abdominal pain (31.4%) of the damage is the most common injuries. Severe lower limb trauma is common in pedestrians and cyclists (20.6% and 11.0%), the average injury severity score was 38.7. Traffic accident mortality through research found that brain trauma 33.58%, 8.09% vertebral trauma, chest trauma, 2.89%, 1.73%, bizarre abdominal trauma. 79.19% axial parts (skull 60% 9.19% thoracic spine 5%, 4% of the pelvis, abdomen, 2%); limbs 20.80%. Alcohol abuse analysis showed that 35% of drivers are deaths due to alcohol abuse.

4 Cause Traffic Accidents

Traffic accident cause analysis model there are many. 1990-1999, to calculate and compare the 13 road traffic safety indicators, Egypt on rural roads accidents, injuries, deaths and casualties predictive models and historical data on the prediction model revisions; using hierarchical regression Non-parametric statistical analysis of the geometric characteristics of rural roads, the accident rate and the relationship between its predictive value. Study various road accidents collection characteristics and provide a quantitative assessment of the feasibility of a simple method. The results show that, although the importance of isolation between the variables and the multi-channel and two-way type, the variable geometry and surface conditions affect the rate of accidents variables are the two most important factors.

Psychological factors in the prediction of traffic accidents, a certain proportion, which predict the reliability of the accident, as the dependent variable of time and unexpected guilt is one of the important research content. Road traffic accident cases and research based on data from the microscopic point of view of its own motor vehicle driver errors and accidents, the establishment of the driver's own errors caused by road traffic accidents due to the model. The model using the improved AHP to determine the driver's own errors led to the accident the main factors affecting the weights, reducing the subjective factor analysis accuracy. The results show that the error in the motor vehicle driver's own road traffic accidents, driver error rate of 50.2% perception, judgment decision error rate 38.9% 10.9% Operating error rate. Vehicle driver perception and judgment of road traffic accidents is less important factors, the findings for motor vehicle driver accident prevention and control provides a scientific basis.

During 1996-1998, the capital of Riyadh in Saudi Arabia occurred 651 serious traffic accidents (accidents caused at least in a personal injury or death) and property damage from 1123 (PDO) accidents) conditional probability and contingency table analysis showed, improper driving behavior, is in Riyadh signalized intersection accidents the main cities, which, running red lights and brake failure is the main cause. An urgent need to review existing intersection geometry and traffic control equipment, also need public education campaigns and enforcement strategies. From 10 British police department for the 1185 fatal vehicle occupant sample study the following conclusions: ① than 65 percent of accidents involving excessive speed driving, the driver alcohol content exceeded, not wearing a seat belt. ② young drivers usually at night in rural areas. Leisure. State driving causing accidents, these accidents and speeding, reckless driving, alcohol and have a great relationship. ③ fewer older drivers involved in traffic accidents, resulting in the death of older drivers tend to be false, erroneous perception, especially in rural areas during the day on the road. In Denmark, there is interdisciplinary specific types of road traffic accidents and accident-depth investigation and analysis of the study, the team head-on collisions, left turn accidents, truck accidents and single vehicle accidents studied. Research data including police reports, the team's investigation of the accident scene and the vehicles involved, and with the road users involved and witnesses to interview. Head collisions and single vehicle accidents are a major factor in excessive speed, drinking and driving and the driving under the influence of illegal drugs. The main factor in the accident left the accident was not paying attention or time left to complete miscarriage of justice. Lack of visual information, speeding is a major factor in truck accidents occur. For all accident injury factor is not wearing a seat belt. Multidisciplinary approach provides a more complete knowledge of the reasons for the accident; this method requires a lot of resources for a common type of accident or serious incident analysis.

5 types of traffic accidents due to traffic accidents often have heterogeneous data, which may cause some degree of potential hazard.

Commonly used cluster analysis to determine the homogeneity of the traffic accident type. First, the heterogeneous traffic accidents dataset into seven categories, namely seven types of traffic accidents. Second, for each type of injury analysis. The class will be based on these results and the complete results of data analysis compared the potential application of a preliminary analysis of clustering methods can reveal hidden relationships. Swedish rural roads using Poisson and negative binomial regression model study found that when estimating the marginal effect of traffic, do not consider the differences between the types of vehicles will lead to loss of important information. If you do not consider the type of vehicle, the accident rate will be reduced; when cars were different types of studies, the accident rate remains unchanged or increased; while the opposite conclusion trucks, indicating an increase in the number of trucks will result in the reduction of accidents. Logistic regression analysis was used to assess the severity of the driver and the relationship between accidents. By analyzing the 1991 to 1996 of 10320 from road traffic accidents found that compared to non-fatal accidents, women in the proportion of fatal accidents than men (odds ratio (OR) = 0.65; 95% confidence interval (95% CI) ,0.53-0 .80). Compared aged <30 years, age ≥ 65 have been injured in traffic accidents significantly increased the probability (OR = 10.87; 95% CI of 4.45-26.54). Among them, the car drivers (OR = 1.85, 95% CI, 1.08-3 .18), moped riders (OR = 3.53; 95% CI, 1.42-8 .78), cycling (OR = 7.72; 95% CI, 2.56-23 .29). Roads outside the city center pedestrian, car drivers, moped drivers and cyclist’s death ratio also increased significantly. Driver's injuries were not using seat belts is also closely linked to the (OR = 13.27; 95% CI ,9.39-18 .74, was mortally wounded; OR = 2.49; 95% CI ,2.17-2 .86, non-fatal bodily injuries , focusing on protecting the weakest road users, under the protection of the law in changing behavior and changing the environment, based on simple interventions may reduce road traffic mortality.

6 traffic accident frequency analysis

Traffic accident frequency analysis is also an important research topic. For traffic accidents of time, often for a longer period of time analyzing the collective (road traffic accidents) and personal (real-time analysis of road traffic accidents) behavior. Through the system reliability theory appropriate adjustments can be largely based on personal approach used to analyze traffic accident frequency. Multi-vehicle accidents is that people study in one direction. Multi-vehicle traffic accident refers to two or more than two traffic accidents between moving objects. With a single vehicle accident is different, not all drivers involved in accidents are responsible for the accident. Road type, speed limit, as well as the number of vehicles involved in accidents are injured in multi-vehicle accident seriousness of the important variables. Fewer parts in motor vehicle fatalities in traffic accidents is also an important global issue. By studying accidents caused the death of regional differences, with targeted interventions. Data from this study from Taiwan police reports, hospital data and vital registration data, and in the pre-hospital deaths and health care at the hospital fatalities for special attention. The results show that the victims of traffic accidents in rural areas to be transported to distant medical centers, rather than the local hospital, resulting in a higher proportion of pre-hospital death. Therefore, specific interventions, such as intelligent emergency medical system, sport helmet and seat belt use, helmet and seat-belt enforcement, and speed control measures should be targeted to rural areas. Simultaneously. Automotive industry in rural traffic accidents rescue teams, emergency medical cooperation between institutions in these areas may reduce the number of deaths. In Zagreb, Croatia, in 1999 to 2000, urban road traffic accident (RTA) for risk analysis, in order to reduce the incidence of injury increases. Through simple binary analysis χ2, odds ratio and 95% confidence interval is used to determine the risk group three results: death, serious and slight injuries. Study found that: There are 528 of the RTA victims, including 260 serious and 213 minor injuries and 55 were injured at the scene and died during transport. More fatal accidents occur at night (OR = 3.78, 95% CI, 2.08-6 .85), urban road intersections (OR = 2.33; 95% CI, 1.30-4 .19), speeding (OR = 2.56; 95% CI, 1.43 -4.61) when. More people in the urban intersection (OR = 5.27; 95% CI of 2.21-12.57) injury rather than death. Driving too fast in urban intersections and poor visibility (OR = 16.15; 95% CI for the 3.901-66.881), the males have a higher death or serious injury of integrated risk.

2.2.2 Study Review

Miaou and Lum road safety information system according to the U.S. (HSIS) occurred in 1985-1987 more than 900 large truck accident data were used Poisson regression model and linear regression model number and road traffic accidents between spatial structures. The results show that the Poisson model is more suitable for the analysis since the number of traffic accidents and road space structures. Kraus and Anderson 1986-1987 according to the California Highway traffic accident statistics to the accident was the dependent variable, taking into account the interaction between the independent variables, using Poisson regression models were established driveway, off road accident rate. Al-Ghamdi occurred in Riyadh under 560 traffic accidents, the accident severity into fatalities and non-fatal accident categories, regression analysis using Logic accident occurred at a time, location, type, driving experience, type of vehicle and the relationship between the severities of accidents. The results show that the reasons for the accident location and severity of the accident has an important influence. Sze and Wong pedestrian accident under the Hong Kong statistics analyzed by Logistic model fatalities and serious injuries influencing factors, indicating that the road environment and the mutual influence of population on the severity of traffic accidents have a significant impact. Kim et al in 1990 in Hawaii accident statistics, classification according KABco accident severity into the death, disability, sexual non-disability injuries, potential injury and property damage only five levels, namely the use of log-linear Logistic model analysis model and the type of incident and the relationship between the severity of accidents. Aguero, Valverde traffic-related factors such as the use (such as a car mileage and road length), socio-economic factors (such as age statistics and population sex ratio) and environmental factors (such as rainfall) predicted the Pennsylvania county accident hazards model. Hadayeghi use of GIS systems such as the City of Toronto planning area evaluated traffic safety. Graham and Glaister using macroscopic models predict an accident under British tutelage pedestrian accidents rank number. Many studies have used the negative binomial model severity of the accident were predicted, including the American scholar Abdel-Aty and Radwan, Shankar, British scholar Quddus and French scholars Amoros and so on. Forecast on road accidents, the number of accidents to zero sections more, many studies using zero accumulation probability model that explains the phenomenon of discrete accident data, fitting better, including the American scholar Shankar, Jinsun Lee, Dominique Lord, etc. .

#### 联系我们

全国免费热线：400-086-0807 咨询QQ：634602927 投稿邮箱：634602927@qq.com |