MOTORCYCLE ACCIDENT MODEL ON THE ROAD SECTION OF HIGHLANDS REGION BY USING GENELARIZED LINEAR MODEL

Transcription

1 International Journal of Civil Engineering and Technology (IJCIET) Volume 8, Issue 10, October 2017, pp , Article ID: IJCIET_08_10_127 Available online at ISSN Print: and ISSN Online: IAEME Publication Scopus Indexed MOTORCYCLE ACCIDENT MODEL ON THE ROAD SECTION OF HIGHLANDS REGION BY USING GENELARIZED LINEAR MODEL Sobri Abusini Department of Mathematics, Brawijaya University, Malang, 65145, Indonesia Lasmini Ambarwati Department of Civil Engineerinng, Brawijaya University, Malang, 65145, Indonesia ABSTRACT There is a high rate of road traffic accidents happening on highways, especially those involving motorcycles. Motorcycle accidents are a major crash occurrence in traffic flow, especially in mixed traffic. However, there are insufficient studies, using a model based on geometric and traffic flow in highlands region, conducted on motorcycle accidents. This study focuses on motorcycle accidents happening in the highlands region of Batu City, East Java Province, Indonesia as a case study, to solve the problem of high rate of accidents involving motorcycles, by employing the Generalized Linear. Based on the accidents data, the number of traffic accidents in the city of Batu is considered to be high. By using accidents data, observation on traffic and road geometry, the model was developed. Using some statistical analysis, it is found that motorcycle accidents model that is influenced by traffic flow, shoulder width and speed, are the best-fit. Therefore the local government should improve some related factors (flow, shoulder width and speed) in order to reduce the number of motorcycle accidents in the city of Batu. Keywords: Motorcycle Accident model, Motorcycle involvement, Generalized Linear, Highlands region. Cite this Article: Sobri Abusini and Lasmini Ambarwati, Motorcycle Accident on the Road Section of Highlands Region by Using Genelarized Linear, International Journal of Civil Engineering and Technology, 8(10), 2017, pp INTRODUCTION Transportation safety is a global problem, It is as much of a social problem as it is a transportation problem [1]. WHO attempted to raise awareness on traffic safety by setting a theme on the World Health Day 2004 Event: Road Safety is no Accident. According to WHO, the rate of road traffic accidents in the world has reached 1.2 million deaths and over editor@iaeme.com

2 Sobri Abusini and Lasmini Ambarwati million injured / disabled per-year. Eighty five percent of the victims of traffic accidents that occurred in developing countries, died, while the number of vehicles that exist in these countries, are only 32% of the total population of vehicles in the world [2]. Transportation is economically inefficient due to the impaired mobility caused by congestion and accidents.. Transportation spending approximately costs 30-40% of the income of an average person. This cost should only amount to 10% of the average income. [3]. The high number of motorcycles on the road have the significant effect of increasing the rate of motorcycle accidents occurring. Batu City as a case study area is a highlands area which has a high number of traffic accidents. The number of traffic accidents in Batu City is explained in Table 1. Motorcycle accidents make up approximately 63% of all traffic accidents. The number of accidents increases each year due to the lack of serious handling of the road safety program. Motorcyclist has a higher risk of getting involved in traffic accidents than drivers of other types of motorized vehicles. The first reason for the higher risk is the instability of motorcycles mainly caused by their design of only being supported by two wheels, and the skill of the motorcyclists. The second reason, is that collisions caused by high speeds have a much more severe effect on motorcycles. Other reasons are because motorcycles lack the equipment needed to protect the rider. Table 1 Data accident in Batu City Type Batu City Accident Die Serious injuries 10 3 Minor Injuries Total of Accidents Source: (4) The impact of motorcycle accidents are financially more harmful than any other kinds of auto accidents. ADB estimated that motor vehicle crashescaused 15 billion USD worth of damages for 10 countries in ASEAN. The highest loss occurs in Indonesia with an estimate of 6.03 billion USD worth of damages per year. Researchers have always focused on motorcycle accidents and developmentof safety programs. A model for motorized vehicle crashes had been developed i.e. Generalized Linear (GLM) [5]. There are Insufficient studies on motorcyclist behavior and driving characteristics on the highlands region, and development of accident model for it. Therefore, the objective of this study is to develop a model for traffic accidents involving motorcycles in the city of Batu. The remaining paper is organized as follows. Accidents model is explained in Section 2, the following sections describe material and methods, clarifying significance of variables influencing motorcycle accidents. Results and Discussions section explains how to describe the variables influencing accidents model in detail. Finally, we present conclusions and recommendations editor@iaeme.com

3 Motorcycle Accident on the Road Section of Highlands Region by Using Generalized Linear 2. ACCIDENTS MODEL A transportation model is a model to illustrate the relationships between land use and transportation by defining mathematical equation model [6;7;8]. for accidents on roads (X-roads) using Generalized Linear [9], the model is explained as follows: = (1) = (2) Where A = average motorcycle accident Q = traffic flow g 1,g 2, = variable geometry α, k, b 1, b 2, = parameter estimation In developing a model for traffic accidents, the particular models of motorcycle accident scenarios are conducted to minimize the number of accidents by using multiple distributions in the statistics [10]. This study tries to understand traffic accidents, especially motorcycle accidents on several roads in the city of Batu. Distribution parameters apply to the highway geometric distribution. The data on road geometrics in Batu City can be seen in Table 2. Table 2. Data Input in Batu City NO Road Acc NL LW SW Median Gradien 80% Speed FLOW 1 Pattimura street Sudirman Street Dewi Sartika Street Surapati Street Raya Mojorejo Street Raya Payung Street Kalilanang Street Raya Punten Street Source: Survey Note: Acc. = accident, NL : number of lane, LW : lane width, SW: Shoulder width The data of 15 roads, are processed by performing several iterations. From analysis using SPSS version 16 and Easy fit 5.2 Standard version [11], it is shown that the best and most significant parameters can be identified from 8 roads in Batu City, as explained in Table MATERIAL AND METHODS 3.1. Distribution Test Before the model of motorcycle accidentsare set up, it is necessary to check the suitability of the distribution of each variable [12]. To determine the shape of the distribution of data and to define most appropriate distribution, analysis is done by using the number of accidents (Acc) as the dependence variable or response variable [13]. To determine the suitability of the distribution, Kolmogorov-Smirnov test was conducted by applying Software Easy fit 5.2 Standard and using the formula: = (3) The distribution tests were conducted on the data of 8 roads involving one response variable (the number of accidents (Acc)) and 5 explanatory variables, i.e., the number of lanes (NL), lane width (LW), road shoulder width (SW), the average speed (S) and traffic volume (F). The results are explained in Table 3 and Table editor@iaeme.com

4 Sobri Abusini and Lasmini Ambarwati Tabel 3. Fitting Results of KS Test No No Distribution Parameters 1 Geometrik p= Logaritmik θ= Neg. Binomial n=1 p= Poisson λ= Source : Result Analysis Table 4 Goodness of Fit-Summary Distribution Kolmogorov Smirnov Statistic 1 Geometric Logarithmic Neg. Binomial Poisson Rank 3.2. Calculation of Kolmogorov-Smirnov Test From the analysis as explained in Table 4, the number of motorcycle accidents (Acc) fits on the Geometric distribution (first order) with the value of KS statistic approximately , Negative Binomial distribution (second order) with the value of KS statistic approximately and Logarithmic distribution (third rank) with KS statistical value of Furthermore, Kolmogorov-Smirnov hypothesis is used to determine a distribution accepted or rejected from a distribution of empirical data using Goodness of Fit Details. Table 5. Goodness of Fit - Details on Geometric [#2] Kolmogorov-Smirnov Sample Size Statistic P-Value Rank α Critical Value Reject? No No No No No Source: Result Analysis Which, H 0 = Data of accident following geometric distribution and H 1 = Data of accident not following geometric distribution. Statistical Test Kolmogorov-Smirnov or D max = Critical Value Kolmogorov-Smirnov 5% or D (5%) = The result is D max <D (5%). It means that H 0 is not rejected (accepted). By considering an error rate of 5%, it is concluded that the accident data (Acc) following the Geometric distribution, with p = From distribution results and graphs of P-P plot, it shows that the best distribution fits on the Geometric distribution. It means that motorcycle accidents data on 8 roads in the city of Batu are following the Geometric distribution, as informed in Figure editor@iaeme.com

5 Motorcycle Accident on the Road Section of Highlands Region by Using Generalized Linear P-P Plot P () P (Empirical) Geometric Figure 1 P-P Plot, Geometric distribution for the City of Batu 3.3. Distribution of Geometric on Kolmogorov-Smirnov Test For more specificality, the following steps perform statistical calculations by using Kolmogorov-Smirnov test against Geometric distribution based on the value of the maximum absolute deviation (D max ), as defined: = (4) Which, D max : the maximum absolute deviation value between F n (x) and F 0 (x) F n (x): the function of cumulative odds observed with n = 8 (the number of observations) = [!"# ] (5) F 0 (x): the cumulative distribution function of Geometric distribution & = = &.&(&) (6) As seen in Table 6, D maks is This value is almost equal to the value obtained using software DmaksEasyfit 5.2 standard. It means that the suitability of the distribution of Kolmogorov-Smirnov test has similar values with the distribution test performed by using the statistical program, Easyfit version 5.2 standard Table 6. Geometric Distribution Calculation In Kolmogorov-Smirnov test x Frek Frek Kum S(x) Fo(x) S(x)-Fo(x) Dmax S(x(i-1))-Fo(x) Dmax Table 7. Summary R R Square Adjusted R Square Std. Error of the Estimate a Predictors: (Constant), Flow, shoulder Width, Speed editor@iaeme.com

6 Sobri Abusini and Lasmini Ambarwati Table 8. Analysis of Variance Sum of Squares df Mean Square F Sig Accident Regression a Residual Total a. Predictors: (Constant), Flow, shoulder Width, Speed b. Dependent Variable: Acc Table 9. Coeffisient 1 Standardized Unstandardized Coefficients Coefficients B Std. Error Beta (Constant) shoulder Width Speed Flow Dependent Variable: Acc A. Estimation of Initial parameters Three stages are conducted to estimate the initial parameters for motorcycle accidents in Batu City. The calculations are explained in Table 6-Table 8. Table 7 explains the determination coefficience of the model (R2) is It means that motorcycle accidents (Acc) are influenced by 94.3% of the variables, i.e. shoulder width, speed and flow. Other variables are only influencing 5.7% on motorcycle accident (Acc). From simultaneous effect (F test) in Table 8, the value of the F test is while the significant number is 0.002, which means that the independence variables, i.e. speed, flow and shoulder width simultaneously effectthenumber of motorcycle accidents (Acc) (dependence variable or response variable). Table 9 explains the effect of each independence variables to dependence variable (response variable). All of independence variables are significantly at 0.05 because the values of all grades of asymptotic having significantly less than It is concluded that motorcycle accidents variable (Acc) is influenced by factors, i.e. flow, speed, and shoulder width. B. Accident Accident model on the road is approached by using Generalized Linear (GLM), defined in equation (1) [9]. Three variables, i.e. flow, speed and shoulder width can be put into the algorithm. It can be defined as follow: ** =+.,-. &../,0 -#0"/+.. 0 (7) Which Acc = the number of motorcycle accidents F = Flow (pcu/hour) pcu = passenger car unit SW= Shoulder Width (m) S= Speed (km/hour). t Sig editor@iaeme.com

7 Motorcycle Accident on the Road Section of Highlands Region by Using Generalized Linear C) Non Linear Regression Analysis To predict non-linear regression model, equation (7) is used to estimate the parameters in developing the model of motorcycle accidents. From Table 10, the model of motorcycle accidents on roads in the city of Batu, are as follows: ** =(.&2& 34..(& &.))567+&.&26 (8) Table 10. Parameter Estimation Parameter Estimate Std. Error 4. RESULTS AND DISCUSSION Lower Bound Upper Bound k 7.02E E E-05 F SW S Comparison of Data Table 11 and Fig.2 explains the estimation of accidents model approaching the observation data with a 95% confidence level. It means that development of accident model is in accordance to the existing condition. Table 11. Data Observation and Estimation in Batu 95% Confidence Interval Roads Acc (Obs) Estimasi Residu Residu MSE Figure 2 Graph Data Observation and Prediction Accident in Batu editor@iaeme.com

8 Sobri Abusini and Lasmini Ambarwati 4.2. Pairwise Test By testing paired samples with t-test, the relationship between the observation data and the results of estimation can be explained, as described in Table 12. Table 12. Paired Samples Statistics Mean N Std. Deviation Std. Error Mean Pair 1 Acc Predicted Values Table 13. Paired Sample Correlation N Correlation Sig. Pair 1 Acc & Predicted Values Table14. Paired Samples Test Paired Differences Mean Std. Deviation Std. Error Mean t df Sig. (2-tailed) Pair 1 Acc - Predicted Values The paired sample statistics in Table 12, explains that the observational data does not differ significantly with the estimated average data (from13.12 into 13.21). Table 13 describes analysis of paired samples correlation, revealing a relationship between observation data and the estimation. The results show that the correlation is very strong (0.992). at the Sig. value (0.000) <α (0.05). This rejects the hypothesis of H 0 : µ Acc. = µ estimates and accepts H 1 : µ Acc. µ estimates. Table 14 about paired sample tests, t-arithmetic (0.158) is less than t-table (8; 0.025) of It means that H 0 is accepted. From comparison of Sig (2-tailed) and α, Sig (2-tailed) approximately > α (0,025). It also means that the observation is similar to the estimation Contribution of Accident Variables Considering the coefficience of each variable from the motorcycle accidents model above, the occurence of motorcycle accidents contributed variably. a) Influence of total volume of vehicles (Flow) on the number of accidents. The model shows that Flow has a significant effect on the rate of motorcycle accidents occuring. It figures that accidents will not occur, if there is no flow on the road. The number of motorcycle accidents will increase regarding to the volume of vehicles on the roads in the city of Batu editor@iaeme.com

9 Motorcycle Accident on the Road Section of Highlands Region by Using Generalized Linear b) Shoulder width influences on the number of accidents. The coefficient of the road shoulder ( ) states that each addition of 1 meter of shoulder width will reduce the number of accidents accuring by approximately e or c) Effects of velocity (speed) to the number of vehicle accidents The coefficient of speed of means that the increase of 1 km/h of speed could cause a motorcycle accident by multiplication e 0,091 or It shows that high speed will tend to increase the number of motorcycle accidents. 5. CONCLUSIONS AND RECOMMENDATIONS 5.1. Conclusions Through the observations and analysis of motorcycle accidents on roads in Batu City, several conclusions are drawn. a) Motorcycle accidents model in the Batu City area. b) The Geometric road factor that's influencing motorcycle accidents is the shoulder width of the road (SW). c) The traffic factors that are effecting motorcycle accidents, are speed and traffic volume. d) From testing the model distribution, the empirical data of this study is in accordance to Geometric distribution 5.2. Recommendations Table 1 Developing a better model.a high number of accidents samples should be considered on traffic. Table 2 Improving the data base by studying further regarding to the identification of blind spots, the cause of traffic accidents and management of traffic accidents handling. Table 3 Further research is needed on the causes of the accidents especially regarding the human aspects, vehicle and environment. REFERENCES [1] Hoel, L. A., Garber, N. J., Sadek, A. W., Transportation infrastructure engineering: A multimodal integration. Cengage Learning, [2] World Health Organisation (WHO), World Report on Road Traffic Injury Prevention (Summary), WHO Library Cataloguing in Publication Data, Genewa, [3] Sulistio H., Role of Government on Road Safety Program, National Symposium of Transportation Department, Malang, [4] Police Department of Batu City, Accident Statistic in Batu City, [5] McCullagh, Nelder, Generalised Linear s, 2 nd Edition. Upper Saddle River, NJ:Prentice Hall, [6] Ortuzar, J. D. D., Willumsen, L. G., ling transport, 2002, Vol. 3. [7] Ofyar Z. Tamin, Planning and Transport ling, Publisher: ITB, Bandung, [8] Russo, F. Comi, A., A modelling System to Link end-consumers and Distribution Logistics, European Transport, 28: 6-19, editor@iaeme.com

10 Sobri Abusini and Lasmini Ambarwati [9] Tony Swan, Robert Gilchrist, Malcolm Bradley, Mike Clarke, Peter Green, Allan Reese, John Hinde, Andrew Stalewski, Carl O Brien, The Glim System, Release 4 Manual, Oxford University Press Inc, New York, [10] Aitkin, M., Anderson, D., Francis. B. and Hindle, J., Statistical ling in GLIM, Clarendon Press, Oxford, England, [11] Ghozali,I., Applications Multivariate Analysis with SPSS Program, the Agency Publisher Diponegoro University, Semarang, [12] Conover, W.J., Practical Nonparametric Statistics, Third Edition, John Wiley & Sons, Inc. Texas Tech University, New York, [13] Zuherman, M., Introduction to Queueing Theory and Stochastic Teletraffic, arxiv preprint arxiv: [14] Ciro Caliendo, Maria Luisa De Guglielmo, Accident Analysis Based on Fixed and Random-Parameters s. International Journal of Civil Engineering and Technology, 8(9), 2017, pp [15] Jasvinder Singh, Mahipal Singh, Anil Baliram Ghubade and Manjinder Singh Analytical Hierarchy Process for Road Accident of Motorcycle in India: A Case Study. International Journal of Mechanical Engineering and Technology, 8(7), 2017, pp editor@iaeme.com