Mode choice model for public transport (ferry, fast ship and aircraft) study case: Kupang-Rote Route

The mode of transportation that travels from Kupang City to Rote Ndao Regency is currently a mode of sea transportation and air transportation. This study aims to find out user information based on its mode choices and model it in the multinomial log function, besides that it’s hoped that this result can be used as a reference in determining policies that improve functions and facilities for mobilizing people, goods and services carried out on the Kupang to Rote.Data analysis was carried out in two ways, namely with descriptive statistical analysis techniques to explain the characteristics of respondents and multinomial logistic regression analysis to create a model of mode transportation selection for Kupang-Rote service. The results of the study showed that users of transportation services from Kupang to Rote, generally men prefer to use ferry boats while women prefer to use fast boats. The characteristics of respondents is age 20-39 years. While the factors that influence the choice of transportation are the influence of luggage, weather and comfort. For the selection model, the probability of a speedboat is 34.65%,the probability of choosing a ferry is 42.96%, and the probability of choosing an aircraft is 22.3%.


Introduction
East Nusa Tenggara Province (NTT) has 427 islands, including Rote Island. Formerly Rote was included in Kupang Regency, but since 2002 it has been established as Rote Ndao District. Rote Ndao itself is Indonesia's southernmost island which borders Australia. Geographical conditions between Rote Ndao District and Kupang City which are separated by the sea cause traffic movement to fulfill the needs of goods and services between islands. The mode of transportation that travels from Kupang City to Rote Ndao Regency is currently a mode of sea transportation and air transportation, whether by plane, ferry boat or fast ship. The main sea ports in Rote Ndao Regency are Baru Beach and Baa. Consideration of transportation selection certainly depends on factors that can affect service user satisfaction. The factors used in the study include income, age, gender, purpose of travel, terminal waiting time, travel time, travel costs, and service level. From the results of his research, the factors that have a significant influence in making modal choices are departure time, travel costs and also reasons for traveling [1]. By adopting the above significant factors accompanied by the addition of several other factors such as distance traveled, luggage and weather influences, then the modeling of transportation mode selection between ferries, speedboats and aircraft serving Kupang departure route to Rote analysis is carried out. This study aims to find out user information based on the choice of mode and model it in multinomial log functions,where to find out the characteristics of the users of the transportation mode for the Kupang -Rote route and find out the factors that influence the choice of transportation modes between ferries, fast boats and planes in the Kupang -Rote route, besides that this result can be used as a reference in determining policies that improve functions and facilities to mobilize people, goods and services carried out on the Kupang to Rote route.

Model of mode selection 2.1 Model definition
In the transportation planning process, one step that must be done is analyzing each relevant data and information as a basis for predicting what will happen in the future. This data and information can be secondary data that obtained from related agencies, or as primary data can be obtained from direct observation on location of research or in reality [4].Primary data obtained from direct observing activities is difficult to analyze and cannot be used as a basis for predictions of incident and results in the future [4]. For necessity of estimation of these results, this realistic data and information needs to be simplified as optimally as possible, without deviating from the intent, purpose, and substance of the relevant data and information.Summarize and simplify the realistic conditions activities are known as modeling activities. Thus, the model can be defined as follows [4]: ➢ The model is a concise representation of the real conditions and tangible forms of design that can explain or represent the real conditions for a particular purpose [2]. ➢ Model is a representation or formalization in certain agreed languages upon from a particular condition [6]. ➢ Model is a main framework or formalization of information / data about collected real conditions to study / analyze the real system [3]. Therefore, the model is a tool that can be used to represent and simplify a reality. The simplification of reality aims to provide a deeper explanation and for the purposes of forecasting.

Utilities
Value of travel satisfaction in using alternative modes of transportation, influenced and related to variables that are considered to have a strong relationship with the behavior of the traveler and the form of the relationship can be seen through the utility functions are as follows [4]: Where : U = The value of the traveler's utility using transportation modes V1 to Vn = Variables that are considered to influence the value of satisfaction using certain modes of transportation F = Mathematical functional relationship The activity of determining and observing the behavior of the traveler through the utility function can be done with two approaches. The approach that used determines the probability choice model. Both approaches are:

Aggregate approach
The aggregate approach is an approach that analyzes the behavior of the traveler as a whole. The aggregate approach can be done in two ways, namely [4]: ➢ Divide the object of observation on several groups that have the characteristics of relatively homogeneous elements (the same). ➢ Aggregate the disaggregate data, which functions for a particular group can be derived from individual utility functions as members of the group.

Disaggregate approach
The disaggregate approach is an approach that analyzes the behavior of individual travelers. In this approach, the value of satisfaction is more realistic because it considers the unobservable elements that occur in the reality. All unobservable elements that occur in the reality, in this approach is represented by an element of error (random) that is random or stochastic, so the model becomes: Where: Um = value of satisfaction function using m mode tm-cm = value of satisfaction function using m mode β1-β3 = value of satisfaction function using m mode en = error factor or stochastic element, that is a random variable that follows a certain distribution form β0 = characteristic constants of alternative satisfaction values, if all variables tm to vm are zero. Forecasting is said to be relatively precise, if the en value is at least close to zero (to a minimum) or en = 0

Sample determination
To determine the amount of samples from a population, can used the Slovin formula is as follows [7]: Where : n = number of samples N = number of population d = percent error rate (d value = 10%)

Population
Population is an area of generalization consisting of objects / subjects that have certain qualities and characteristics determined by researchers to be studied and then drawn conclusions [5]. The population in this study consisted of two populations that were distinguished based on the location of the study, namely the population of transportation service users from Kupang to Rote. Population data for users of transportation services both from Kupang is shown in Table 1 and Table 2.

Sample
Sample is a part of the amount and characteristics owned by the population [4]. Based on existing population data, the sample calculation is carried out as followsequation 5.
In each study area there are 3 points of the research location, in Kupang consists of Tenau Harbor, Bolok Port and Eltari Airport while in Rote it consists of Ba'a Port, New Beach Harbor and D.C Saudale Airport. Then, the amount of respondents for each study area was divided into 3 parts. One hundred respondents from Kupang were divided into 3 locations, each location received 33.33 or equal to 34 respondents. However, in this study 35 respondents were selected for each research location, likewise withRote area the questionnaire distributed for each location in Rote was 35 questionnaires. So that the questionnaires distributed to respondents in Kupang and in Rote amounted to 150 questionnaires (can be seen in Table 3)

Data collection techniques
Primary data is obtained from the results of questionnaire surveys in the field. Distribution of questionnaires was given to respondents who were ferry service users, speed boats and aircraft. The questionnaire survey was conducted at the port and airport waiting room.For aircraft departing from Kupang the questionnaire survey was carried out in the waiting room of Eltari International Airport. Ferry boats were also carried out in the waiting room and also in ships at Bolok Port (which departs from Kupang).Meanwhile, a survey of fast boat service users was carried out in the waiting room and in ships in Tenau Port (passengers from Kupang to Rote). The form of the survey form questions includes two things. First, questions are made to determine the characteristics of the users of transportation services and the characteristics of the transportation system of the three modes of transportation. Second, questions are made to find out the respondents' preferences towards alternative modes of transportation selection with several types of changes in conditions offered.

Primary data
Primary data is data that can be used directly from the original source. The primary data in this study were sourced from the results of a survey of needs for ferry passengers, speedboats and planes from Kupang and from Rote. The data that will be obtained from the survey results in each access mode are as follows: ➢ Characteristics of Respondents The characteristics of the data collected are as follows age, work, travel purpose, reasons for choosing transportation,travel frequency, monthly income, distance and travel time from the place of origin to the airport or port, distance and travel time from the airport or port to the destination and estimated total transportation costs in traveling ➢ Service user preferences for service attributes collected are as follows Cost (X1), Travel time (X2), Luggage (X3), Weather (X4) and Comfort (X5)

Secondary data
Secondary data is data obtained from relevant agencies, in this case from PT. ASDP (River Lake and Crossing

Data analysis techniques
Survey data that has been collected is then processed so it can be used as input data in the analysis process using the IBM SPSS version 23 application program. Data analysis is done in two ways, namely descriptive statistical analysis techniques to explain the characteristics of respondents and multinomial logistic regression analysis for make a model of choosing the transportation modes for Kupang-Rote service route. After the equation model is obtained, the statistics are tested to find out whether the mode selection equation is feasible to use. Then do the probability calculation of each mode selection for Kupang to Rote route and the opposite route. The probability calculation of each mode is in accordance with equation 6,8,10 which in this study was carried out on ferries, fast ships and aircraft. The modal choice probability calculation formula is as follows:  Table 4 below:

Simultaneous test
Based on the results of the analysis seen from Table 5, this test serves to determine the significant effect of independent variables on transportation mode selection decisions (dependent variable). The Examination Hypothesis is as follows: -H0 means that the independent variable does not affect the dependent variable -H1 means that at least one independent variable has a significant effect on the dependent variable  Table 5, it can be seen that the chi-square value is calculated at 214.019 while the chi-square table (0.05; 62) is 81.3376. This means that the chi-square count is greater than the chi-square table and also seen the p value of 0.000 <0.05. Thus the H0 hypothesis is rejected and states that at least one independent variable has a significant effect on the dependent variable or in other words, the overall independent variable influences the choice of transportation mode selection.

Partial test
The Likelihood Ratio Tests table in this test shows the contribution or influence of each independent variable on the transportation mode selection model. From the partial test results, it can be seen that the probability value smaller than alpha 5% is the variable influence of luggage (0.012), the influence of weather (0.038) and the influence of comfort (0.011). Then the variable that contributes to the model is the luggage, weather and comfort variables while the cost and travel time variable does not contribute to the model because it is not significant (p> 0.05).

Model feasibility test
In  The results of Model feasibility test or Goodness of Fit can be seen in Table 7 From the table it is known that the value of D (68,323) ≤ Xtable (105,264), then H0 is accepted and the significance value is 0,091 above the alpha value of 0,05. It can be concluded that the model matches the empirical data or the multinomial logit model is suitable for use because a ood model is a model that accepts the null hypothesis.

Model equations of Kupang-rote selection mode
The logit function generated based on Table 8 parameter estimation is as follows: Logit 1 U(fast ship-aircraft) = 0,437+ 0,054X1-0,741X2-1,951X3-1,603X4 + 2,197X5 Logit 2 U(ferry-aircraft) = 0,652 + 0,129X1-1,079X2+2,125X3+0,575X4-2,059X5 The variables that influence the model in logit function 1 and 2 are those that have a significance value of p <0.05. In logit 1 function, the comparison between fast ships and aircraft shows that the variables that affect the modal selection model for the Kupang-Rote route are luggage variables and weather variables. Significant p value of each variable is p of luggage variable 0.013 <0.05 and the weather variable significant p value 0.031 <0.05. The luggage variable has a lower influence on the selection of fast boats compared to aircraft with a coefficient value of -1.951 and significant at p (0.013) <0.05. The weather variable in the logit 1 function gives the effect of choosing a fast ship lower than the aircraft. Coefficient value of Weather variable -1.603 and significant at p <0.05, means that when the weather is inadequate for sea transportation to be used, respondents will tend to choose a plane.
In logit 2 function, the comparison between ferry and aircraft shows that the variables that affect the modal selection model for the Kupang-Rote route are luggage variables and convenience variables. The significant value of p of luggage variable is 0.009 <0.05 and significant p of the convenience variable is 0.006 <0.05. The luggage variable has a higher influence on the selection of ferry than the aircraft with a coefficient of 2.125 and a probability of 0.006 (p <0.05). The convenience variable gives the effect of choosing a ferry boat lower than the aircraft with a coefficient value of -2.059 and a probability of 0.006 (p <0.05). Based on logit 1 and 2 equation, the multinomial logistic regression model is obtained as follows: Based on the calculation of the transportation modeprobability for the service route from Kupang to Rote, the results are obtained as above. Fast ship probability is 34.65%, ferry probability is 42.96% and aircraft probability is 22.38%. It is clearly known that of the three modes that have the largest probability of selecting transportation mode is the ferry with a percentage of probability 42.96%.

Discussion of ferry, fast ships and aircraft users characteristics (Kupang-rote routes)
Based on the characteristics of ferry users analysis, fast ships and aircraft for both service routes shows the magnitude of movement in each of these modes. The form of the modal selection model for a service route as was done in this study, can provide the same or different results for each route. This is due to the presence of several influencing factors. Based on the results of research found that the factors group that influence the mode selection of Kupang-Rote are as follows: ➢ Characteristics of travelers In this factor group, all variables relate to the individual of the traveler. The variables referred to include factors of gender, age, average monthly income and type of work. Based on the analysis results, several conclusions were drawn regarding the characteristics of ferry transportation, fast ship and aircraft users for both service routes and described as follows: a. The effect of gender factors on the transportation characteristics users for both service routes describes different conditions. For the Kupang-Rote service route, male travelers generally prefer to use ferries while women prefer to use fast ships. b. The influence of age factors on the characteristics of transportation users for both service routes illustrates the same conditions. The Kupang-Rote service route describes the characteristics of the travel agent dominated by transportation users in the age range of 20 to 39 years. This is because the age range is a productive age range that travels for work. c. Most of the travelers who traveled from Kupang were earning less than one million rupiah per month. Most respondents who travel by ferry are respondents with income less than one million rupiah per month. Respondents who earn more than four million rupiah per month prefer to use airplanes, while average respondents that use fast ship earn less than one million rupiah, but the percentage is not as many as the ferry respondents. Judging from the model of equations that are formed, the factors that influence the choice of transportation modes from Kupang are luggage, weather and comfort factors. The luggage factor influences the choice of mode between ferries, fast ship and aircraft because each transportation mode has a different luggage capacity limit for passenger. Luggage capacity on aircraft is limited to each person is 15 kg while ferries and fast boats do not have luggage capacity limits for each passenger. However, ferries have the advantage of being able to carry vehicles while fast boats are not possible. This causes the luggage factor gives an influence on the mode transportation choice.
The next factors that gives effect to the probability model of mode selection is weather. The weather factor influences the choice of mode because the available transportation modes to connect between Rote and Kupang are sea transportation modes and air transportation modes. In certain months such as January-February and June-August, weather influences do not support the use of sea transportation due to high winds and high sea waves. Travelers who travel in these months will generally switch modes from sea transportation to air transportation (aircraft). This also makes travelers think that the comfort level of the aircraft is higher than the other transportation modes.
Based on the influence of luggage, weather and comfort factors on the model of modal selection for Kupang-Rote route, a large probability for each mode of transportation is obtained. The highest probability value for the Kupang-Rote route is on a ferry with a percentage of 42.96%.

Conclusions
1. Users of transportation services from Kupang in the direction of Rote, in general men prefer to use ferry women prefer to use fast boats. The characteristics of respondents for travel is ages period of 20-39 years.
The average user of this service comes from Govermert employee. There are people who travel for work and events for families. When traveling, on average you have to travel more than 10 km to get to the airport and port. 2. Factors that influence the choice of transportation modes between ferries, fast boats and planes both Kupang-Rote are luggage, weather and comfort. Luggage is the reason for ferries and fast boats to be more dominant. While climate and comfort are the reasons for choosing a plane.Kupang-Rote service route 3. Utility function model of mode selection between fast ships, ferries and aircraft are as follows: Ufast shipaircraft = 0,437 + 0,054X1 -0,741X2 -1,951X3 -1,603X4 + 2,197X5 Uferryaircraft = 0,652 + 0,129X1-1,079X2 + 2,125X3 + 0,575X4 -2,059X5 Logistic multinomial model: Fast ship= The probability of fast ship selection is 34.65% Ferries= The probability of ferry selecting is 42.96% Plane= The probability of plane selecting is 22.38%