Makalah Time Series Tugas Besar-2
Makalah Time Series Tugas Besar-2
Makalah Time Series Tugas Besar-2
Intervention Model for The Impact Of Lapindo Hot Mud Flow Disaster To Vehicles Volume In Toll-Road
Winda Eka Febriana 1, Aizeh Maulidina 2, Heri Kuswanto3, Jerry Dwi, T. P4
result is expected to be used as additional information for the relevant parties in taking the various policies concerning the operational processes in Waru-Gempol toll gate. II. BASIC TEORIES Intervention Model Intervention model is a model used at external events beyond the estimates as well as internal events are expected to affect the forecast variables and internal events that affect these variables. This analysis has the main purpose to measure the magnitude and duration of effect of intervention on a time series. A common form of intervention model (Wei, 1990) is:
Yt = s ( B) B b X t + Nt r ( B)
Abstract - The hot mud flow disaster from PT. Lapindo Brantas, Porong, caused the Porong toll-road flooded and has to be closed. This closure is affecting to the vehicle volume that passing the Surabaya-Gempol toll-road, especially in WaruGempol toll-road. To determine the occurrence of large disaster effects happened, it is necessary to analyze the time series with step function intervention model. The data used was the vehicle volume data from January 2000 to December 2007. Step function intervention model, depicts the major effects caused by the Lapindo hot mud flow disaster to the volume of vehicles passing through the toll Gempol-Porong. By the disaster, the volume of vehicles that pass through the toll road vehicles decreased by 387.181 units of one to three months after the occurrence of mud flow (June to August 2006). The effect of this decrease is still happen until the end of the research, so the effect that caused by the Lapindo hot mud flow is the permanent effect. Index TermsHot mud flow, intervention model, step function.
I. INTRODUCTION Surabaya-Gempol toll road is a toll road along the 42 miles that connected the city of Surabaya by Gempol. On 29 May 2006, there was a rare phenomenon that is the emergence of hot mud mixed with gas from one of the houses of residents in Porong, south of Sidoarjo. Mudflow came from the mining area (Wells Banjar Panji 1) owned by PT. Lapindo Brantas in Sidoarjo. Hot mud flow that increasingly extends to drown in the surrounding settlements. As one of the Surabaya-Malang access link, SurabayaGempol toll road also negatively affected by the Lapindo hot mudflow. Porong-Gempol toll road on kilometer 38 and 39 were flooded by the mud that cannot be bypassed by a vehicle. This toll road was closed several times because of stagnant mud and finally had to total close on 26 November 2006. Closure of Porong-Gempol toll road has resulted in fewer vehicles passing by vehicle users prefer an alternative route to go to Malang or Surabaya. Based on data from the PT. Jasa Marga Persero, previously the number of vehicles that use toll roads carrying about 39.000 vehicles/day. In this research, the problems to be studied is the impact of Lapindo hot mud flow on the volume of vehicles in WaruGempol toll road which is a closed system of motorway toll road in Surabaya-Gempol. The analysis will be used is time series method with intervention model step function. The
In general, there are 2 types of intervention variables (Wei, 1990), there are: 1. Interventions that have occurred since time T and so on. In this case the intervention variable is a valuable step function as follows (Wei, 1990):
X t = St
(T )
0 , t < T = 1, t T
2. Interventions that occur only at time T only and does not continue at the next time. At this condition, this intervention variable is a pulse intervention function as follows (Wei, 1990):
X t = Pt
(T )
1, t = T = 0, t T
III.
METHODS
3.1 Data Sources Data used in this research is secondary data about the volume of vehicles on the highway closed Waru-Gempol per month starting in January 2000 to December 2007. Data obtained from the PT. Jasa Marga Persero, Tbk SurabayaGempol branch. 3.2 Intervention Variables In this research, there are events that allegedly intervention effect on the volume of vehicles on the highway WaruGempol. The incident was a case of the mudflow, which started the observation at t-77 to be exact on May 29, 2006 (t
2 77). Although the intervention occurred in May 2006, but the effect was only felt in the months thereafter, and until now still continues, so that the intervention variable in this research is a step function with the following equation: 1, t 77 S = 0, t < 77 . 3.3 Methods of Analysis 1. Determining ARIMA model to the data before the intervention. 2. Determining the intervention model, with this steps: a. Identify the order of the intervention model b. Estimate parameters and significance testing of the intervention model parameters. c. Diagnostic check of residual intervention model. d. Calculate and interpret the effects of intervention model. IV. 4.1 Descriptive Data
Bulan Januari Februari Maret April Mei Juni Juli Agustus Septembe r Oktober Nopember Desember TOTAL
steps in the modeling data before the intervention were as follows: 4.2.2 Identify the ARIMA Model Model identification can be done by looking at time series plots, plots of ACF and PACF plots. Identification is done to see whether the data stationary in mean and variance. Plot time series of data the total volume of vehicles from January 2000 to December 2007 can be seen in figure 4.1 below:
Ti me S er i es P l ot of vol ume
2400000 2200000 2000000 v olume 1800000 1600000 1400000 1200000 1000000 1 10 20 30 40 50 Index 60 70 80 90
Figure 4.1 Time Series Plot of Data Vehicles Volume at January, 2000-Desember,2007
Ti me S eri es P l ot of pre-i ntervensi
2300000 2200000 2100000 pre-int erv ensi 2000000 1900000 1800000 1700000 1600000 1500000 1400000 1 8 16 24 32 40 Index 48 56 64 72
Table 4.1 Descriptive Statistics Volume Vehicle from January 2000December 2007 by Month
Mean 1.772.945 1.597.845 1.784.753 1.702.290 1.789.026 1.758.324 1.883.798 1.809.973 1.783.718 1.822.283 1.749.035 1.788.805 1.770.233
St. Dev 228.007 231.920 215.447 275.110 243.109 233.864 251.828 240.466 286.342 242.537 171.679 259.446 236.849
Minim 1.449.059 1.215.761 1.414.361 1.155.665 1.388.281 1.450.522 1.555.737 1.546.912 1.429.110 1.459.959 1.494.176 1.385.497 1.155.665
Maks 2.100.547 1.925.511 2.063.087 2.024.122 2.098.842 2.089.452 2.283.541 2.189.908 2.232.119 2.107.304 2.015.163 2.185.450 2.283.541
Figure 4.2 Time Series Plot of Data Vehicles Volume at January, 2000-April,2006
From Table 4.1 it can be seen that the average volume of vehicles on the highway Waru-Gempol that most occur in July even as many as 1,883,798 vehicles in a month. Meanwhile, the smallest average volume of vehicles occurred in February that as many as 1,597,845 vehicles. The largest standard deviations of vehicles volume occur in September in the amount of 286 342 and the smallest standard deviations occur in November in the amount of 171 679. Maximum vehicles volume occurred in July in the amount of 2,283,541 vehicles and vehicles volume minimum occurred in April in the amount of 1,155,665. .4.2 ARIMA Model of the Volume of Vehicle before Intervention Modeling vehicles volume data using ARIMA models performed before the incident intervention. The model will be used as a noise model to form models of intervention. The
Based on Figure 4.1 can be seen that the data is there overall incidence of intervention step function of the Lapindo hot mud disaster (S) from T = 77. In Figure 4.2, the data volume of vehicles on the highway Waru-Gempol from January 2000 to April 2006 (data prior to the intervention) had a trend to rise. This proves that the data is not stationary in the mean because the average is not constant and is still influenced by time. ACF and PACF Plot for the data volume of vehicles prior to the intervention are shown in Figure 4.2 and 4.3.
Autocor relati on Function for N t
(with 5% significance limits for the autocorrelations) 1,0 0,8 0,6 0,4 0,2 0,0 -0,2 -0,4 -0,6 -0,8 -1,0 1 5 10 15 20 25 Lag 30 35 40 45
3
Residual ARIMA(1,1,0)(0,1,0)12 ARIMA (0,1,1)(0,1,0)12 0.8158 0.7618 0.9209 0.9328 0.9804 0.9852 0.8713 0.8068 0.8834 0.8435 0.9038 0.8576 0.9667 0.9470 0.9881 0.9835 0.9848 0.9819 0.8598 0.8365
Lag 6 12 18 24 30 36 42 48 54 60
Because the data is not stationary in mean and variance, then the data must be reduced in the lag-1. In addition, the data are also not stationary in a seasonal pattern, as based on the ACF plot in Figure 4.3, shows that data to 12 and multiples thereof, tend to have a different mean, where the value of autocorrelation lag out of bounds at the 12 and 24. Therefore, data must be in-differencing and lag-1 lag-12.
Autocor r el ati on Functi on for pr e-i nt-di ff-di ff1 2
(with 5% significance limits for the autocorrelations) 1,0 0,8 0,6 A ut ocorre lat ion 0,4 0,2 0,0 -0,2 -0,4 -0,6 -0,8 -1,0 1 5 10 15 20 25 30 35 Lag 40 45 50 55 60
Based Table. 4.3 can be seen that the residuals of both models to meet the assumption of white noise because of all the Pvalue is larger than is 0.05, so it can proceed to the next test.
Table 4.4 Normality Test for Residual of the Model before Intervention
Table 4.4 shows that the value Pvalue for KolmogorovSmirnov test on the residuals from the two ARIMA models larger than . Then it can be concluded that the residuals for both models with normal distribution at significance level of 5% error. Further analysis is to select the best model from the models.
Table 4.5 Criteria of Model Selection before Intervention
(a)
P artial Autocor rel ati on Function for pre-i nt-di ff-diff1 2
(with 5% significance limits for the partial autocorrelations) 1,0 0,8 Par t ial A ut ocor relat ion 0,6 0,4 0,2 0,0 -0,2 -0,4 -0,6 -0,8 -1,0 1 5 10 15 20 25 30 35 La g 40 45 50 55 60
Based on Table 4.5, note that the value of AIC and SBC model ARIMA (1,1,0) (0,1,0)12 smaller than the value of AIC and SBC model ARIMA (0,1,1) (0,1,0)12. It can be concluded that the model ARIMA (1,1,0) (0,1,0)12 is the best. Based on this model, can be written into the following equation:
Z t = Z t 1 - 0.36494 Z t 1 0.36494 Z t 2 + Z t 12 Z t 13 0.36494 Z t 13 0.36494 Z t 14 + a t
(b) Figure 4.5 ACF and PACF Plot of data before the intervention after the differencing
Based on Figure 4.5 (a and b) can be determined for alleged order ARIMA model of vehicle volume data. Based on the ACF and PACF plot, allegations of ARIMA models, including: ARIMA (0,1,1) (0,1,0)12 and ARIMA (1,1,0) (0,1,0)12. Furthermore, both models will be tested, to determine the best model.
Table 4.2 Estimated Parameters for the Model ARIMA (0,1,1) (0,1,0)12
Parameter 1
From Table 4.2 can be seen that the parameter for the model ARIMA (0,1,1) (0,1,0) 12 and ARIMA (1,1,0) (0,1,0) 12 is significant at = 5%, so it can continue on the next test.
Table 4.3 PValue Residual from Both ARIMA Models for Data before Intervention
| |
| | |
4
20 0 0.00000 | . | . |
Based on Figure 4.6, can know the order alleged intervention model is b = 1, s = 0, r = 0.
Table 4.6 Estimated Parameters for the Intervention Model
Parameters
NUM1
From the equation above, can be seen that the Lapindo hot mud flow disaster has reduced the number of vehicles passing the highway Porong many as 420,016 units of vehicles within one to three months after the event (in June, July, and August 2006). This is because the mud flow is quite close to the location of Porong toll road, so road users are afraid to pass through toll roads Porong and prefer an alternative way to go to Malang or Surabaya. V. CONCLUSION Based on the discussions that have been obtained in the previous chapter, it can be concluded as follows: 1. Intervention models appropriate for the data volume of vehicles on the highway Waru-Gempol because of the Lapindo mudflow is as follows:
Yt = -420016.4 St + at (1 + 0.3758 B )(1 B ) (1 B12 )
From the Table. 4.6, can be seen that all the parameters for the models of intervention with the value b = 1, r = 0, and s = 0 is significant at 5% error rate, so it can proceed to the next test.
Table 4.7 Pvalue Residual Intervention Model
Lag 6 12 18 24 30 36 42 48 54 60
Pvalue 0.4946 0.1140 0.1727 0.2363 0.3847 0.4206 0.5218 0.7335 0.8646 0.9151
Based on Table. 4.7 can be seen that the residuals of intervention models to meet the assumption of white noise because of all the Pvalue is larger than (0.05), so it can proceed to the next test.
Table 4.8 Normality Test for Residual of Intervention Model
In this intervention model, events (interventions) that have an impact on the volume of vehicles that pass through Waru-Gempol toll road is the Lapindo hot mud flow (St) for t 77 which is a step function intervention model. 2. The calculation and explanation of the large and the length of the effect caused by the Lapindo mudflow disaster are as follows: Lapindo hot mudflow disaster has reduced the volume of vehicles on the highway Porong as much as 387,181 units of vehicles one to three months after the occurrence of mud flow (June to August 2006). References [1] [2] [3]
Box, G.E.P., Jenkins, G.M., and Reissel, G.C., 1994. Time Series Analysis Forecasting and Control, 3rd edition. Englewood Cliffs : Prentice Hall. Cryer, J.D., 1986. Time Series Analysis. Boston : Publishing Company. Daniel, W.W., 1989. Statistika Nonparametrik Terapan. Jakarta : PT. Gramedia. [4] Jasa Marga, 1987. Jalan Tol Surabaya-Gempol. PT. Jasa Marga (Persero), Tbk cabang Surabaya-Gempol. Makridakis, S., Wheelwright, S.C., and McGee, V.E., 1999. Jilid 1 Edisi Kedua, Terjemahan Ir. Hari Suminto. Metode dan Aplikasi Peramalan. Jakarta : Bina Rupa Aksara. Widyaningsing, Retno., 2008. Intervention Model To Evaluate The Impact of Lapindo Mud Flow Disaster To Vehicle Volume in Turnpike, Final Project S1 Statistika ITS Surabaya (tidak dipublikasikan).
Pvalue >0,1500
Table 4.8 shows that the value Pvalue for KolmogorovSmirnov test of residuals from intervention model is greater than , so we can conclude that the residual model follows a normal distribution at a significance level of 5% error. The model of the data volume of vehicles on the highway after the intervention can be written in the following equation: Yt = 0 St + Nt Nt = at
[5] [6]
(1 1B )(1 B ) (1 B12 )
at Yt = -420016.4 St + (1 + 0.3758 B )(1 B ) (1 B12 ) Based on models of intervention that has been obtained, then the calculation of the Lapindo hot mud flow effect on the number of vehicles is as follows: - Effects on the period T (May 2006) Yt = -420016.4 St = -420.016 - Effects on the period T+1, T+2, T+3 (June, July, and August 2006) Yt + k = -420016.4 St + k = -420.016 , untuk k=1,2,3