Motorcycle accident modelling for the formulation of motorist safety action programs in Kupang City

Kupang city had to deal with transport safety issues where the total number of traffic accidents in 2017 increased by 51.05% compared to 2011. This study aims to describe and analyze the factors that affect motorcycle accidents through modelling by the method of Generalized Linear Models with Genstat and SPSS programm. Based on the results of the police report data summary, accidents most often occur on Sunday (15.95%); the time span at 18:00 to 23:59 pm (45.94%); type of hit the front (35.68%); victims died (12.19%), male sex (80.71%); age range 18-25 years (37.20%); with the level of education at the level of Higher Education (42.38%); worked in the private sector (37.75%); do not have a driver's license (75.43%). The results of the analysis of survey data obtained equation modelling accident MCA = 0,0006713* Flow 0.000275*exp [0.2144 LaneWidth ShoulderWidth_Sidewalk 1.952 2.026 MedianRoad + 0.2139 Speed + 0.0513 Access]. Modelling results showed that the addition of total lane width decrease the number of motorcycle accidents by 10% per year, pavement by 6% per year, median by 13% per year, traffic flow by 3% per year, speed by 14% per year, access road by 2% per year. It is therefore recommended a program of action in the form of additional elements of the road medians and pavement on a road segment that the accident rate can be reduced.


Introduction
Community needs for efficient and affordable transportation modes have resulted in a growing number of motorbikes in Indonesia. The use of motorcycles in Indonesia is very popular because the price is relatively cheap, so it can be reached by most people because of the use of fuel and operational costs that are quite economical. The rapid growth of motorized vehicles is not followed by the growth of road infrastructure and good traffic procedures by motorists, causing negative impacts such as accidents and air pollution. According to data from the Kupang City Traffic Unit (SATLANTAS), the number of traffic accidents (Lakalantas) in 2017 increased by 51.05% compared to 2011. The number of victims due to traffic accidents in 2017 also increased by 46.49% compared to 2011.
This research is part of an effort to find solutions to reduce motorcycle accidents that continue to increase every year in the City of Kupang. By knowing the significance of these accident factors, an action program can be prepared which is expected to be more effective and on target in reducing the number of motorcycle accidents through modelling by the method of Generalized Linear Models with Genstat and SPSS programm. This research is limited by several things as follows: 1. Identification of the road accidents causes studied only covers traffic factors, namely the flow, speed and composition of the vehicle, and the road geometric factors. 2. The area studied is the arterial road network and collectors in the administrative area of Kupang City. 3

Vehicle flow
Flow is the number of vehicles passing a road in a certain time unit. Traffic flow is usually measured by placing a counter at the current place which wants to be known, or it can be done manually. Calculations can be made for vehicles that move in one direction or two directions. The general formula used to calculate traffic flow is [1]: Traffic flows consist of various types of vehicles such as passenger vehicles, trucks, buses and motorcycles. In order to facilitate the calculation, the size of the traffic volume of various types of vehicles is stated in the Passenger Car Unit by multiplying certain factors or the Passenger Car Equivalent .

Speed
Speed is the ability of a vehicle to travel a certain distance per unit of time, speed is usually expressed in meters/seconds or kilometers/hour (kph). The basic equation for vehicle speed is [1]: This study uses 85-percentile in speed data processing. 85-Percentile meaning a traffic speed where 85% of drivers drive their vehicles on the road without being affected by lower traffic speeds or bad weather.

Generalized linear method (GLM)
Generalized Linear Models (GLM) used for create a model which observation data (response variable) is not normally distributed. The general equation used in the GLM method is [6]: Where : FK = The frequency of traffic accidents that will be predicted Xi, Yj = explanatory variables (i = 1,2,3, ..; j = 1,2,3.) α, β = variable coefficient

Stages of analysis of the GLM method
Correlation test, used for determining the attachment of relationships between fellow explanatory variables, which are expressed with correlation coefficients. Fellow explanatory variables cannot be correlated with each other. The parameter value/correlation coefficient (r) is: > 0.9 -1.0 means having a very strong relationship; > 0.7 -0.9 means having a strong relationship; > 0.5 -0.7 means having a moderate relationship; > 0.3 -0.5 means having a weak relationship; > 0,0 -0,3 means having a very weak relationship; Negative or positive signs only indicate the character of relationship between variables. If positive, the otherwise. Two variables have a unidirectional relationship, and Univariate analysis, conducted to determine the effect of each explanatory variable individually, and at the same time check the level of significance of the response variable (number of motorcycle accidents). The initial description of the contribution of explanatory variables can be known from the results of this analysis.
Multivariate analysis, where the quantitative effects and significance of several explanatory variables together on the response variable will be examined. This is necessary because the cause of an accident is a combination of several factors.

Stages of collecting data
In this study, data collection was carried out using the survey method at the research location. The survey conducted was as follows: 1. Preliminary survey Preliminary survey in the form of a traffic flow survey on the review road. The preliminary survey aims to determine the peak hour of the vehicle flow which will be used as the basis for determining the time of the next traffic survey.

Geometric Road Survey
Geometric road survey aims to identify geometric elements of the road including the width of the road lane, the width of the road shoulder, the calculation of the number of access and the presence of the sidewalk.

Traffic Survey
This survey aims to obtain traffic flow data on roads and the proportion of motorbikes to the total traffic flow. The survey was conducted by calculating the number of vehicles classified according to vehicle type.

Research Stages
The steps taken are as follows:

Result of accident data descriptive analysis
The characteristics of motorcycle accidents on the observed road are the number of accident victims based on gender, severity, type of work, level of education, type of crash, time of occurrence (day and range of hours), the proportion of age and ownership of Driving License (SIM) by the perpetrator and victim.

Fig. 1. Motorcycle accident victims based on gender
The diagram in Figure 1 shows that the average motorcycle accident between 2011 and 2017 is more common among male riders with a total of 2567 people (80.02%). This can be caused by the number of male vehicle users are larger than female riders, so the risk of accidents experienced by male riders is greater than female riders.  In Figure 3 it can be seen that the front crash type has a more dominant number than other types of crash. The highest front hit occurred in 2015 where there were 126 cases (26.47%). This type of crash can be caused by the absence of a median of the road or is affected by the disorderly behavior of the driver. This can be caused by the fact that during this time the road flow conditions are more flexible so that undisciplined motorists tend to drive vehicles at high speeds which increases the risk of accidents. Another factor that can also be a cause is at these times driver is susceptible to being in an unfit condition such as exhaustion or drowsiness.

Fig. 6. Motorcycle accidents based on type of work
The relationship between the level of motorcycle accidents and the type of work is shown in Figure 6 where most of the drivers who had accidents worked in the private sector (37.04%). The highest number of private professions that experienced accidents occurred in 2017, which amounted to 248 people (21.20%). This can be affected by higher mobility of private workers compared to motorists who have non-private jobs.

Fig.7. Motorcycle accidents based on age proportion
Based on the diagram in Figure 7 it can be seen that the age most frequently involved in accidents is the age range of 18-25 years (37.87%). This can be caused by the fact that most motor vehicle users are in that age range, so the likelihood of accidents at that age range is the biggest and also the relatively young age factor can effect unstable emotions and disorderly driving behavior. The highest number of accidents in the age range of 18-25 occurred in 2017, as many as 274 people (22.50%).

Fig. 8. Motorcycle accidents based on education level
Based on the diagram in Figure 8 it can be seen that the drivers who are most often involved in accidents are motorists with College education level (42.39%). The highest number of accident victims with College education levels occurred in 2015, as many as 114 people (26.76%). This can be caused by the number of motorcycle riders in Kupang dominated by college students scattered in various areas of Kupang City so that the risk of accidents on motorists with this level of education is greater than motorists with other levels of education.  Table 1 shows the recapitulation of motorcycle accidents and ownership of driving license (SIM) from 2011 to 2017 on the observed road section Piet A. Tallo  42  10  32  3  Frans Seda  76  23  53  4  Eltari  30  8  22  5  Timor Raya  258  74  184  6  Ahmad Yani  21  3  18  7  Soekarno  8  5  3  8  Siliwangi  2  2  0  9 Pahlawan 49 12 37 10 Yos Sudarso 18 6 12

Geometric and traffic survey results on reviewed roads
The survey is carried out for 10 (ten) days (5 working days and 5 weekend days). Table 2 shows the recapitulation of the results of the geometric survey and traffic on the road section that observed.

Correlation test
Correlation test, was conducted to determine the attachment of relationships between fellow explanatory variables, which are expressed with correlation coefficients. Based on the correlation test results in Table  3

Univariate test
The Univariate Test is a test carried out on individual explanatory variables on the response variable. The aim of Univariate analysis is to find out individual explanatory contributions to response variables. Only variables that are significant variables will then be included in the multivariate analysis.

Multivariate test
Multivariate analysis was conducted to determine the quantitative effects of explanatory variables simultaneously on the response variable. Based on the data in Table 5, the variable Width Lane with a coefficient value of 0.001, the variable Shoulder and Sidewalk Presence with a coefficient of 0.001, median variable with a coefficient of 0.001, a current variable with a coefficient of 0.029, a variable speed with a coefficient of 0.001, and an access variable with coefficient value 0.005 has fulfilled the significance requirement smaller than 0.05 so that it can be included in the modelling results.

Modelling results
The results of multivariate analysis with the GLM method, obtained the final model for the prediction of motorcycle accidents in the city of Kupang are as follows.

Model limitation
Modelling generally has limitations related to the suitability of the model with the characteristics of the existing location. This needs attention for all parties who will use a model on the location (sample) with different characteristics than when the model was made. The model formed in this study is suitable for traffic characteristics and geometric elements of the road segments as follows: The total range of lane width : 6.

Result of accident model interpretation
Based on the results of modelling using the Genstat program there are several variables that affect the number of motorcycle accidents on arterial roads in Kupang City. These variables can increase and reduce the percentage of motorcycle accidents on the road segments studied. The influence of each variable on the number of accidents can be explained as follows.