Travel pattern analysis and freight trip generation modeling of textile commodities. Case study: Thamrin City Shopping Center, Jakarta

One of the sectors supporting the country’s economic development is the commercial sector. With the increasing population, the demand for consumer goods will increase, impacting trip generation. Thus, it is necessary to conduct a study to predict the trip generation of freight transport in shopping centers, especially in DKI Jakarta, with textile stores as trip generators. This research aims to determine the travel pattern of freight transport and freight trip generation models produced and attracted at the Thamrin City shopping center and to analyze the factors influencing freight trip generation. The research was conducted using a quantitative research method with a cross-sectional. Data collection used a questionnaire given to 105 merchants. Data analysis used linear regression methods with the number of freight trip production and attraction as the dependent variable. The independent variables used are the number of employees and gross floor area. The study results show that the delivery of goods using expedition services is sent to the distribution center before being sent to the destination. While shipping goods using private vehicles, sent directly to the destination location. The gross floor area significantly influences the trips generation. Transportation planners can use this research to estimate the number of trip generation in shopping center areas in the future.


Introduction
Technology has become a key driver of economic, regional, and urban growth in this modern era.Economic growth is closely related to the ability to produce and distribute goods and services.One of the sectors supporting economic development in a country is the commercial sector [1].Over time the number of people in urban areas will increase.The larger the population, the greater the demand for consumer goods [2].This growing number of people encourages the development of a transportation system to handle the entire flow of goods.However, through the development of this transportation system, problems such as congestion can arise.
One of the areas in Jakarta that often faces congestion problems is the Tanah Abang area.This area is known as the largest textile trading center in Southeast Asia.Using land transportation such as trucks as the main mode of transportation for goods movement also causes this congestion.Based on the results of the Indonesia Infrastructure Initiative (2014) study shows that the average number of freight transportation trips is 75.3% using road mode.This congestion causes longer travel times and additional transportation costs that can reduce the productivity of trade activities [4].It is necessary to plan the freight transportation system to solve this problem by estimating the freight trip generation.Through this estimation, the number of freight trips generated and attracted from the Thamrin City shopping center and the model of the number of freight trips generated or attracted from the Thamrin

Literature Review
Trip generation is one of the stages of transportation planning to estimate the number of movements generated and attracted from an area.Holguin-Veras et al. (2013) explained that in freight transportation planning, there are two approaches that can be taken to estimate the generation that occurs.The two approaches are Freight Generation modeling and Freight Trip Generation modeling.Freight Generation (FG) is the generation of commodities transported by vehicles.FG modeling refers to the production and attraction of commodities measured in tonnage or volume (m 3 ).In comparison, Freight Trip Generation (FTG) generates vehicle trips generated in transporting commodities in freight generation.FG is related to the business size; the more significant the industry scale, the more commodity moves into or out of the business location.Meanwhile, FTG is not only influenced by the size of the item transported but also the size of the mode of transportation that transports these commodities.Thus, an increase in FG does not show a linear relationship with an increase in FTG.Commodity generation can increase without necessarily increasing the amount of trip generation.Industry can increase FG/FTG by increasing the size of commodities to be shipped and changes in the size, type, and mode of transportation [5].
In modeling freight generation (FG) and freight trip generation (FTG), the relationship between the number of vehicle trips and the amount of cargo with land use factors is analyzed.Based on previous research, several factors affect FG and FTG are raw materials, production goods [6], land area [7], number of employees [7]- [14], distance to market [9], [10], and type of industry [10], [12], [13].In addition, several methods are used for this modeling, such as linear regression analysis [8], [10], [11], [15], ross-classification analysis [12], and multiple classification analysis [9], [12].Linear regression and multiple classification analyses are suggested for FG/FTG modeling due to their simplicity in modeling and presentation of results [12].In addition, linear regression analysis is the most frequently used method in estimating freight trip generation [8].However, in conducting regression analysis or OLS regression modeling, it is necessary to evaluate the assumptions used because it will affect the quality and accuracy of the modeling.

Study methodology and data collection
The research was conducted on July 18, 2022, with textile commodity merchants at the Thamrin City shopping center as an object study.Sampling was carried out using the simple random sampling method.Two types of variables are used, namely, independent, and dependent variables.The independent variables used in this study are store area and number of employees.Meanwhile, the dependent variables are freight trip production (FTP) and freight trip attraction (FTA).From the sampling results, 105 textile commodity merchants and information related to company characteristics and freight shipping were obtained as respondents.Statistical calculations of the survey results are shown in Table 1.Before the data is subjected to regression analysis, it is necessary to identify outliers.The outlier identification process is carried out using the boxplot method in the SPSS application.Table 2 shows the comparison results before and after outlier removal.The calculation results show a change in the average value and standard deviation.The standard deviation becomes smaller than the mean for NE and GFA variables.However, for the FTP and FTA variables, after outlier removal, the standard deviation is larger than the means; this indicates that individual data points are far from the mean value or the data is widely dispersed [16].

Freight trip generation modeling
This study used the linear regression method with the model form shown in equation ( 1) for simple linear regression and equation (2) for multiple linear regression.According to [11], in modeling, three possible cases occur, namely (a) statistically significant constants and coefficients, (b) only statistically significant coefficients, and (c) only statistically significant constants.It is also mentioned that modeling with constants is recommended except for statistically insignificant constants.
The normality test results show that the data is not normally distributed, and if used for linear regression, the results can be inaccurate.Data transformation is needed into SQRT (x) or square root so the data becomes normally distributed.Before testing linear regression analysis, a correlation test is carried out to measure the relationship between the dependent and independent variables.The calculation results show that both NE and GFA variables are positively correlated for FTP and FTA models, as shown in Table 3. Linear regression modeling was performed with a combination of two and one variable.The best model is determined after all models are tested for linear regression assumptions.One of the assumption tests that need to be fulfilled is the t-test to partially assess the level of significance between the dependent and independent variables.The calculation results shown in Table 4, show that the multiple variable models with intercept do not significantly affect FTP and FTA.The model without intercept shows a strong correlation between variables, so it cannot be used together in modeling.In the NE-based model, it is found that the model with intercept does not have a significant effect on the FTP and FTA models, as shown in Table 5.Thus, since the model's results with intercept have no significant impact, a model without intercept can be used [11].Both without intercept models have a substantial effect on FTG and FTA.In the calculation of the GFA-based model shown in Table 6, it is found that the model with intercept does not fulfill the t-test for the FTA model.However, for the FTG, the model with intercept meets the t-test.Thus, for the GFA-based models, the FTG equation uses a model with intercept, and FTA uses a model without intercept.Based on the t-test results, the equation model with multiple variables does not meet the t-test requirements, so it is unnecessary to do the f-test since the model that meets the f-test is only one variable.Of the four equations that meet the requirements of the t-test, heteroscedasticity testing is carried out to test the occurrence of inequality of variance from the residuals of one observation to another.The conclusion reached is invalid if heteroscedasticity occurs in the equation [17].From the calculation results shown in Table 7, it is known that for the FTP model, both equations fulfill the heteroscedasticity test.Meanwhile, for the FTA equation model, the heteroscedasticity test requirement is only met by the equation with the GFA variable.In modeling freight trip production (FTP), two models were obtained.The best model is selected based on the coefficient of determination and the number of independent variables in the equation.The greater the coefficient of determination and fewer independent variables in the equation, the better the modeling.However, in FTP modeling, it is impossible to compare the coefficient of determination for selecting the best model due to differences in the type of equation, namely the first equation without intercept and the second with intercept.The best modeling is chosen based on the presence or absence of constants in the equation.In the FTP model, the equation with intercept is chosen because if the constant value of the equation is still significant, it is more advisable to use the equation [11].From the equation, it is known that the GFA variable is used to explain the relationship between land use factors and the number of freight trips generated from Thamrin City.The GFA variable is also obtained for FTA modeling to explain the relationship between land use factors and the number of freight trips generated from Thamrin City.The equation obtained in Table 8 is the result of linear regression from the data transformation of the SQRT (x) variable or square root so that in its use, it is necessary to include GFA data in the form of square roots and the amount of FTP or FTA in the form of square roots.In the FTP model, 0.102GFA indicates the relationship between the GFA and FTP variables has a linear component with a coefficient of 0.102, or every one unit increase in GFA will lead to a 0.102 increase in FTP.The constant -0.124 indicates the value added to the relationship between FTP and GFA.This negative constant value of 0.124 estimates the amount of FTP when the value of GFA is 0. However, since the amount of storage area value (GFA) will never be zero, this negative constant value can be ignored.The constant value can provide a meaningful interpretation if the value of the dependent variable is equal to zero within the range in the sample [18].The R 2 value in the FTP equation is 0.212, which means that the GFA variable affects FTP by 21.2%, and other factors influence the rest.In the FTA equation, the R2 value is 0.738, which means that the GFA variable affects FTA by 73.8%, and other factors influence the rest.In addition, the RMSE value in the FTP and FTA equations are 0.374 and 0.183, respectively.Based on these values, it is known that the average prediction error of the model against the actual data is 0.374 for FTP and 0.183 for FTA.In addition, it can also be seen that the GFA variable is a better predictor variable for modeling FTA because the RMSE value is smaller than the FTP RMSE.

Freight travel pattern at Thamrin City shopping center
The travel pattern of freight transport at Thamrin City Shopping Center is divided into trip production and trip attraction.Most of the delivery of goods from Thamrin City is done using expedition services.The types of vehicles used by expedition services are vans and trucks, as shown in Figure 1.Delivery from Thamrin City to the distribution center generally uses a van with an average load occupancy of 70%, then for delivery from the distribution center to another distribution center or the final location using a truck with an average load occupancy of 100%.Delivery to the final location can also use other transportation, such as trains, ships, and planes, depending on the sellers' demand.There are expedition companies that pass through two distribution centers in delivering goods.The first distribution center collects all goods from the area around Tanah Abang.After that, the goods are sent to the second distribution center to be sorted to the destination.For expedition companies that only pass through one distribution center, the goods are sorted directly at the distribution center and grouped according to the destination of the goods.Most goods are sent to the Jakarta, Bogor, Depok, Tangerang, and Bekasi areas.The time required for the goods to arrive at their destination is around 2-3 days.The process of delivering goods to Thamrin City is divided into two types: delivery using expedition services and privately owned vehicles, as shown in Figure 2. Delivery of goods using expedition services uses several kinds of vehicles from the initial location to the distribution center, such as trucks, planes, trains, and ships.The trip from the distribution center to Thamrin City usually uses a van-type car.Expedition companies are also divided into shipments that pass through two or one distribution center.Generally, the first distribution center is to collect all goods that come from the same area.After that, the goods are sent to the second distribution center for sorting goods in the Tanah Abang or Thamrin City area.Delivery of goods using privately owned vehicles is divided into three types of vehicles, namely small, medium, and large trucks.Generally, delivery of goods using privately owned vehicles is sent directly to Thamrin City without stopping elsewhere.Most goods are delivered from the Jakarta, Bogor, Depok, Tangerang, and Bekasi areas.

Conclusions
Based on the research that has been done, the area of the store is the land use factor that affects the freight trip generation in the Thamrin City shopping center.The best multiple linear regression model to predict freight trip generation in the Thamrin City shopping center uses the store area variable (GFA).The model equation for freight trip production is FTP 1/2 = 0.102GFA 1/2 -0.124 with a coefficient of determination of 0.212.Then, the model equation for freight trip attraction is FTA 1/2 = 0.024GFA 1/2  with a coefficient of determination 0.738.
The pattern of freight transport in the Thamrin City shopping center is divided into two, namely, trip production and trip attraction.In shipping goods, most goods are sent using expedition services using vans to the distribution center before being sent to the final consumer using trucks, trains, ships, or planes.In the process of receiving goods, goods can be delivered from the location of origin using private vehicles or expedition service vehicles.Delivery by personal vehicles uses small, medium, and large truck vehicle types.Meanwhile, delivery by expedition services uses truck vehicles from the location of origin to the distribution center and vans from the distribution center to Thamrin City.Most delivery destinations and origin of goods received are Jabodetabek, with the time required to arrive at their destination around 2-3 days.

Figure 1 .
Figure 1.Schematic of trip production chain.

Table 1 .
Summary statistics of dependent and independent variables.

Table 2 .
Comparison of statistical calculations before and after outlier removal.
2are not comparable for models with and without intercept.

Table 6 .
Area based FTG models.

Table 8 .
FTP and FTA regression model.Highlightes cells represents the preferred models; 2) R 2 are not comparable for models with and without intercept.