European Research Studies Journal

Volume XXIII, Issue 3, 2020

pp. 446-459
The Decision Tree Approach for the Choice of Freight
Transport Mode: The Shippers’ Perspective in Terms of
Seaport Hinterland Connections
Submitted 11/03/20, 1st revision 21/04/20, 2nd revision 26/05/20, accepted 07/06/20

Izabela Kotowska1, Marta Mańkowska2, Michał Pluciński3

Abstract:

Purpose: Current research in the area of transport decisions indicates that the key factors
decisive for the mode choice are the cost and the time of transport. The complexity of
behaviours and preferences of cargo shippers as well as the diversity of supply chain
configurations, along with unavailability of an appropriate dataset hinder reliable
forecasting the demand for transport and planning its development by means of quantitative
methods. The aim of this article is to identify the factors that affect the decisions on mode
choice by cargo shippers, based on data obtained by means of a qualitative method.
Design/Methodology/Approach: The decision tree methodology was used in the analysis of
the research study. To analyse the decision tree on the basis of C4.5. algorithm, the authors
applied the J48 module of the WEKA 3.8.4. software.
Findings: The research has shown that the major attributes in selecting transport modes by
cargo shippers, taking into account access to three modes of transport to the seaports
hinterland, are consignment size and time pressure, then owning or having access to barge
terminals by cargo shippers, and the annual volume of cargoes generated by them.
Practical Implications: The results of the analysis can be useful for managers of supply
chain making decisions regarding the choice of transport route.
Originality/Value: The developed decision tree model provides cargo shippers with a
possibility of choosing three transport modes to carry cargoes to/from the seaports: road,
rail, and inland shipping, which constitutes supplementation and expansion of the studies
completed so far, which usually took into account only rail and road transport.

Keywords: Decision tree model, transport mode choice, seaport hinterland transportation.
JEL codes: M15, M20.
Paper type: Research article.

Funding:
The project is financed within the framework of the program of the Minister of Science and
Higher Education under the name "Regional Excellence Initiative" in the years 2019 - 2022;
project number 001/RID/2018/19; the amount of financing PLN 10,684,000.00

1
Corresponding author, Faculty of Engineering and Economics of Transport, Maritime
University of Szczecin, Poland, e-mail: [email protected]
2
Institute of Management, University of Szczecin, Poland,
e-mail: [email protected]
3
Institute of Management, University of Szczecin, Poland,
e-mail: [email protected]
 Iz. Kotowska, M. Mańkowska, M. Pluciński

447

1. Introduction

Transportation modelling is an integral part of forecasting the demand for transport

(Samimi et al., 2012). Mode choice is the third out of four main stages of
transportation modelling which covers trip generation, trip distribution, mode
choice, and assignment (Ortuzar and Williumsen, 2006). With regard to seaports and
their connections with the hinterland, Halim et al. (2016) indicate that availability
and quality of port-hinterland connectivity plays a major role in choosing such ports
by cargo shippers. The mode structure as well as the quality of port-hinterland
connectivity also affect decisions taken by cargo shippers with regard to selecting
the transport mode to/from seaports. Consequently, the strategic decisions made by
cargo shippers and providers of transport and logistics services with regard to
selection of hinterland transport mode, location of distribution centres, and
connections between the distribution centres and transport infrastructure have an
impact on the hinterland transport structure, and thus strongly influence the seaport
connectivity.

The current research studies on the issues of transport decisions mostly apply the
optimisation approach and models of discrete choice based on big data analyses. At
the same time, the academic literature points out to the complexity of behaviours
and preferences of cargo shippers, as well as the diversity of supply chains
configurations, along with unavailability of appropriate datasets, which hinder
reliable forecasting the demand for transport and planning its development by means
of optimisation approaches and discrete choice models (Cunningham, 1982; De Jong
et al. 2004). In this context, the qualitative multi-criteria decision tree methods based
on data obtained from direct research and various decision tree models developed on
their basis are deemed to be very easy to construct and interpret, and to have
considerable cognitive value and practical usefulness (Samimi et al., 2012). Still,
they are rarely applied in researching the decisions of cargo shippers regarding mode
choice in transport to/from seaports.

The aim of this article is to identify the factors that affect the decisions on choosing
transport modes by cargo shippers located in the seaports hinterland, based on the
data obtained by means of a qualitative method (in-depth interviews among shippers
located in the hinterland of the Polish seaports of Szczecin and Świnoujście). As the
developed decision tree model applies the data derived from direct research, the
model to a larger extent covers complex behaviours and diverse preferences of cargo
shippers, which have a key impact on transport decisions made by them.

The research results are presented in the form of a decision tree. The developed
decision tree model provides cargo shippers with a possibility of choosing three
modes of transport to carry cargoes to/from the seaports: road, rail and inland
shipping, which constitutes supplementation and expansion of the studies completed
so far and taking into account only rail and road transport.
 The Decision Tree Approach for the Choice of Freight Transport Mode:
The Shippers’ Perspective in Terms of Seaport Hinterland Connections
448

2. Literature Review

A considerable part of research studies completed so far in the area of port-

hinterland connectivity focused on developing the systematics and conceptualisation
of port-hinterland distribution systems (Van den Berg and De Langen, 2011;
Rodrigue and Notteboom, 2010; Notteboom and Rodrigue, 2017; Chislov et al.,
2019). The research studies on transport decisions made by cargo shippers, also
regarding haulage to/from seaports, apply mainly quantitative methods, such as the
optimisation approach, and discrete choice models taking into account
predominantly the cost and time aspects (Thore and Iannone, 2012; Iannone, 2012;
Guand Lam, 2013; Nam and Win, 2014; Ferrari et al., 2011; Wang et al., 2016;
Tundys et al., 2018).

However, Samimi et al. (2012) point out that the optimisation approaches require
large amounts of data and are not easy to formulate and solve in practice, whereas
discrete choice models have certain unavoidable statistical assumptions such as
linear property of utility function and pre-defined structures (e.g. probability
distributions), which to a certain degree make it impossible to reliably estimate
mode choice behaviours. Therefore, qualitative and multi-criteria methods are more
and more often applied to study transport decisions and mode choice behaviour.

A more in-depth research study in this respect, applying the qualitative approach to a
larger extent, was carried out by Halim et al. (2016) who developed a strategic
model of a network of distribution of cargoes between the port and its hinterland,
taking into account the preferences of cargo shippers and the structures of the port-
hinterland logistics systems (a strategic model for port-hinterland freight distribution
networks).

Also, the multi-criteria methods are gaining popularity, including the methodology
of decision trees, due to their simplicity and considerable cognitive value. The
decision tree methodology is to a larger extent applied in the research studies
regarding the mode choice in passenger transport (Oral and Tecim, 2013; Janssens et
al., 2006; Rasouli and Timmermans, 2014; Tang et al., 2015). However, the
decision tree methodology is rarely used in research studies in the area of freight
transport mode choice and the issue of port-hinterland connectivity (port-hinterland
oriented freight network models). Simultaneously, the usefulness of the decision tree
methodology in the analyses of cargo flows was confirmed by the studies carried out
by Thill and Venkitasubramanian (2015) to explain the nature of inter-port
competition in three dimensions, space, commodity types and shipment values.

Interesting studies applying the decision tree methodology in transport decisions

made by cargo shippers were also carried out by Samimi et al. (2012) in the USA.
The authors took into account attributes such as origin, destination, mode of
transportation, type, value, weight, and volume of the commodity. The studies
included two modes of transport or their combinations truck only, rail, or rail
 Iz. Kotowska, M. Mańkowska, M. Pluciński

449

intermodal. Using various decision tree models, the authors proved that there is a
merely 9% chance of choosing rail instead of road transport. The authors have also
pointed out that long-distance, heavy and containerised cargoes gravitate more to
rail transport, and consignment weight is the most important attribute that influences
the decisions on selecting a transport mode.

The research studies described in this article cover three hinterland transport modes:
rail transport, road transport, and inland shipping, which makes it possible to verify
and expand the studies completed so far, using multi-criteria methods (Samimi et
al., 2012). Thus, the research results presented in this article supplement the
theoretical knowledge regarding the scope of choice from among the three transport
modes (rail, road, barge) by cargo shippers located in the seaports hinterland, with
the use of the decision tree methodology.

3. Methodology

The decision tree methodology is one of the most popular decision support methods
(Quinlan, 1990). A decision tree is a structured tree with a root node from which
decision nodes branch out. Each decision node has one branch coming out of the
root node or a higher level decision node, and two or more branches leading to lower
level decision nodes or a leaf node. A leaf node is the last node which constitutes the
class label, i.e. the final decision result.

In order to develop a decision tree showing the decisions of cargo shippers making
their choices regarding transport modes to be used to carry cargoes to/from the
seaports, we applied Quinlan’s (1990) algorithm C4.5, which is an extension of the
basic algorithm ID3 (Dai and Ji, 2014). The advantage of algorithm C4.5 compared
to algorithm ID3 is a possibility of creating a decision tree based on attributes whose
values do not have to be binary, moreover, it applies the pruning method, i.e.
pruning during the construction of trees to avoid over-fitting (Li and Hu, 2008). The
J48 module of the WEKA 3.8.4. software programme was used in order to develop
the decision tree.

The first stage of the research study was developing a database containing the
factors that are decisive for selecting a transport mode by cargo shippers. The
database was developed on the basis of the direct research study involving cargo
shippers located in the hinterland of the seaports of Szczecin and Świnoujście
(Poland), which was carried out in the form of standardised direct interviews in
2017. The purpose of the research was to specify the potential demand for inland
shipping to/from the seaports of Szczecin and Świnoujście, as an alternative to road
and rail transport, in view of the planned upgrading of the waterway leading to the
seaports (the Oder Waterway). The Oder Waterway is now practically not used as a
hinterland transport mode to/from the said seaports due to its unsatisfactory
technical parameters. Consequently, the completed research studies also helped to
 The Decision Tree Approach for the Choice of Freight Transport Mode:
The Shippers’ Perspective in Terms of Seaport Hinterland Connections
450

verify the existing knowledge in the area of competitiveness factors of hinterland

modes of transport in land-sea transport chains.

The direct interviews were held among 18 maritime exporters and 22 maritime
importers. The interviewees were identified by means of the targeted selection
method, and they concomitantly met the following conditions:

1) running business activity in the catchment area of the seaports in Szczecin

and Świnoujście, i.e. in the provinces of Lubuskie, Wielkopolskie,
Dolnośląskie, Opolskie and Śląskie (the analysis did not include cargo
shippers from the Zachodniopomorskie due to the small distance to the
seaports (<250km), which makes their cargoes naturally gravitate to road
transport)
2) location within a distance of no more than 50 km from the route of the
Oder Waterway, so that all the entities have a potential access to the three
modes of hinterland transport, rail, road, and inland shipping
3) generating annual cargo shipments to/from seaports at the level of 10,000
tonnes or more.

The cargo shippers were identified on the basis of the official data obtained from the
Department of the Analytical Centre of the Tax Administration Chamber in Warsaw
(DACTAC CAAC, data for 2016). The interviews were held with representatives of
all the business entities that exported or imported more than 100,000 tonnes of cargo
per year (7 exporters and 7 importers) as well as 26 entities that generated transport
volumes within the range from 10,000 to 100,000 tonnes per year (11 exporters and
15 importers). These included both primary cargo shippers (manufacturing
companies that export their products or import raw materials/ components) and
secondary cargo shippers (intermediaries) who imports goods for other entities or
export their products. The group of entities covered by the study was representative
for the potential of the hinterland of the analysed seaports in Szczecin and
Świnoujście. The study involved all the entities that determined their transport needs
to exceed 100,000 tonnes of cargo per year, and over 70% of the entities that
generated annual cargo volumes from 10,000 to 100,000 tonnes to be carried to/from
the said seaports.

The obtained responses were applied in creating a database containing 47 instances

(seven cargo shippers make use of two transport modes depending on the adopted
values of attributes). The developed database included the following attributes and
values corresponding to them:

@attribute annual_volume {>100.000,10.000-100.000}

@attribute distance {>600,400-600,<400}
@attribute consignment {>1500,250-1500,<250}
@attribute time_preasure {yes,no}
@attribute barge_port {yes,no}
 Iz. Kotowska, M. Mańkowska, M. Pluciński

451

@attribute producer {yes,no}

@attribute transport {rail,barge,road}.

The database was the source of data for the developed decision tree. In the WEKA
software, the “training set” option was applied, which is used to create descriptive
models in the case of having a database containing all attribute values (Brownlee,
2014).

The analysis was supplemented with additional factors affecting the mode choice,
which resulted from specific external and internal determinants of a given cargo
shipper. The factors were classified in three groups, technical, economic and
organisational, and assigned to the individual transport modes.

4. Results

The decision tree developed on the basis of the obtained data is presented in Figure
1. There were 44 correctly classified instances, which accounted for 93.617% of all.
The research results processed by means of the decision tree method indicating that
the basic attributes in choosing the transport mode by cargo shippers, in the case of
having access to the three modes of transport between the hinterland and the
seaports (rail, road, barge), are as follows: consignment size, time pressure,
possessing or access to barge terminals by cargo shippers, and annual volume of
generated cargoes.

The detailed determinants of the mode choice to be made by cargo shippers in the
seaport hinterland, identified during the primary research study, are presented in
Table 1. Their impacts depend on the volume of a single consignment (small < 250
tonnes, medium 250–1500 tonnes, large > 1500 tonnes), additionally in the case of
medium consignments that do not need fast delivery, their annual volumes (medium
10–100 thousand tonnes or large > 100 thousand tonnes). The analysis has shown
that most cargo shippers transporting their cargoes in consignments exceeding 1500
tonnes available themselves of rail transport. The factors decisive for choosing rail
transport include:

1. considerable carriage distance (>300 km),

2. considerable annual volume of homogeneous cargoes, regardless of their
kind: bulk cargo, general cargo, or containerised cargo (from several dozen
thousand tonnes to several million tonnes),
3. one-off consignments at the level from 1.5 to 2.3 thousand tonnes.

The research study has shown that rail transport was the optimal transport mode for
dry bulk cargoes vulnerable to damage or shrinkage during transshipment (e.g.
brittleness, powder consistency). A limited number of transshipment operations in
the case of direct haulage makes it possible to preserve the appropriate quality of the
commodity.
 The Decision Tree Approach for the Choice of Freight Transport Mode:
The Shippers’ Perspective in Terms of Seaport Hinterland Connections
452

Figure 1. Decision tree for transport mode choice to be made by cargo shippers in
the seaports hinterland, generated in the Weka 3.8.4 programme, with the use of the
J48 algorithm

Source: Own work.

Cargo shippers that generate more than 100,000 tonnes of cargo per year are
inclined to choose rail transport due to the preferential treatment of this customer
group by rail operators. According to the interviewed cargo shippers, rail transport is
fast, cheap, safe and reliable.

The factors affecting the choice of rail haulage in this group of cargo shippers and
some cargo shippers that generate annual cargo volumes at the level of 10–100
thousand tonnes included the infrastructural conditions such as a direct access to the
rail transport infrastructure via a railway siding on the premises and its capability
(provided already at the stage of construction of the industrial plant) to handle rail
transport (e.g. having railway turntables in place), or not having own storage space
(using specialised rail wagons to store cargoes on the premises).

The organisational factors that facilitate choosing rail transport by cargo shippers
include:

1. possibility of transporting, within a short time, a consignment which can be

directly loaded onto a coaster vessel in a seaport or which, upon placing in a
storage yard, may be loaded via an indirect system onto a panamax vessel,
and be taken out of the port,
2. possibility of transporting, within a short time, a consignment of several tens
of thousand tonnes of cargoes from a seaport to the cargo shipper’s premises
in the port’s hinterland,
 Iz. Kotowska, M. Mańkowska, M. Pluciński

453

Table 1. Factors determining the mode choice by cargo shippers in the seaport
hinterland
Determinants Rail transport Inland shipping Road transport
- adapting the premises to - having an own inland port - adapting the premises to
infrastruc rail transport handling - a need to use intermodal road transport handling
tural - not having own transport due to the lack of own - not having own storage
specialised storage space railway siding space
Tech - specific features of the - cargoes not vulnerable to - sensitive cargoes
physical
nical cargoes that require damage in the course of requiring controlled
and
minimisation of numerous transshipment temperatures
chemical
transshipment operations operations
features
- consignment sizes that enable
of cargo
involvement of rail haulage
- short pre-carriage time of - unattractive rail freight - a need for fast
a consignment to the transport offer for the cargo movement of cargoes
seaport to be directly shippers who declare a need for to/from a seaport
transshipped onto a coaster transport of large, single - a need for transporting
- short pre-carriage time of consignments, but not small or medium
a consignment to the frequently consignments on an
seaport to be indirectly - a cargo shipper’s positive irregular basis
transshipped onto a experience regarding inland - dispersed activities of
panamax vessel shipping used in other divisions the cargo shipper
- short on-carriage time of of the company - considerable dispersion
Organisational cargoes from the port, - short distance to the seaport across the hinterland of
which occur occasionally (100 km) customers of cargo
in large maritime - perceiving inland shipping as shippers acting as
consignments the safest transport mode by intermediaries
- making use of a shippers of some specific kinds - considerable diversity of
comprehensive service of cargoes kinds of cargoes handled
offered by an intermodal by any given cargo
operator shipper acting as an
- capital ties between the intermediary
cargo shipper and the rail
operator
- attractive haulage rates - low freight rates and a low - a too small consignment
for cargo shippers risk of losses resulting from (below 1500 tonnes),
generating large annual frozen capital in case of short unattractive to a rail or
cargo volumes freight distances barge operator
- connection between the - no concerns among cargo - a high unit value of a
time of transporting the shippers generating large consignment
cargo to the seaport and annual cargo volumes about - savings resulting from
payment made by the negative effects of modal not having to keep own
contracting party diversification of transport in storage space
(importer) the context of contracts with - cargoes imported on
- concerns of cargo rail freight operators DDP terms
shippers who annually - providing a possibility of - cargoes of unusual size
Economic generate medium cargo transport for cargoes which due and weight, stored in
volumes, fearing that rail to their low value are not bonded areas, where it is
transport rates can increase suitable for rail or road the final customers, not
as a result of modal transport the cargo shipper being
diversification of freight - making occasional deliveries the importer, who are
transport of project cargoes responsible for
- reliability of long-term, transporting the goods to
fixed rate freight service the seaport’s hinterland
- minimising the cost of
frozen capital and a low
risk of incurring penalties
to be paid to customers as
a result of downtime
Source: Own work.
 The Decision Tree Approach for the Choice of Freight Transport Mode:
The Shippers’ Perspective in Terms of Seaport Hinterland Connections
454

3. capital ties between a given cargo shipper and a rail operator or an operator
in the seaport, who does not have an inland shipping infrastructure,
a need to use the services of a specialised intermodal operator who will take
over the whole organisation of the transport process (taking the cargo over
from the cargo shipper’s premises and transporting it to the seaport/
intermodal terminal).

The economic factors that facilitate choosing rail transport result from, among other
things:

1. advantageous rail freight rates offered by rail operators to cargo shippers

who generate considerable annual cargo volumes,
2. attractive terms and conditions of freight carriage are binding throughout the
whole period (e.g. a year) for which the contract was concluded (including
the winter period),
3. concerns of cargo shippers who generate smaller cargo volumes, fearing that
attempts to diversify the transport solutions (e.g. via partial use of inland
shipping) will contribute to increased freight rates that may be offered to
them by rail operators,
4. minimising the frozen capital by cargo shippers and limiting the risk of
penalties to be paid to contracting parties due to rail transport capabilities of
carrying considerable cargo volumes within a very short time (payments
made by importers on the day following the day on which the vessel with
cargo leaves the port of departure) and the low risk of downtime on the
transport route (no traffic jams).

The factors have contributed to achieving by rail haulage a competitive advantage

over other transport modes.

Inland shipping is perceived as an opportunity mainly by those cargo shippers who

make deliveries in medium and large consignments – above 250 tonnes, and none of
the shippers has declared that inland shipping may be the only mode of transport to
carry cargoes to/from the seaports. This results mainly from the low reliability of
delivery attributed to inland shipping, while it is considered a necessary condition
for shifting cargoes from other modes of transport (Kotowska et al., 2018). The
factor that affects the reliability of delivery by inland shipping is navigation closed
periods when deliveries are not made.

The carriers who indicated inland shipping as the dominating mode of transport have
a direct access to the infrastructure (possibility of carrying cargoes directly between
the seaport and the shipper’s premises), which in view of competitive haulage rates
(in relation to rail haulage) translates into considerable financial gains.
Simultaneously, such cargo shippers are not under time pressure in connection with
a longer time of inland transport in relation to rail haulage or they are located
relatively close to the seaport, so rail operators are unable to achieve an advantage in
 Iz. Kotowska, M. Mańkowska, M. Pluciński

455

terms of transport time. Also, the shippers who generate the largest annual freight
volumes are not concerned that making use of inland shipping services will
contribute to being offered worse transport terms and conditions (increased cost
rates) by rail transport operators. Inland shipping would be readily used by the cargo
shippers who are not the priority customers to rail transport operators. These are
mainly the shippers who do not generate full train consignments and the ones who
generate whole train consignments, but not very frequently (e.g. once a month).

These cargo shippers often have to wait for a long time to have any wagons provided
for loading, they cannot count on short delivery times or preferential rates, therefore
they are inclined to search for alternative solutions. The price advantage of inland
shipping may enable trading (and consequently transporting) cargoes whose haulage
with other transport modes, due to their very low value, would not be economically
viable (e.g. sawmill waste). Inland shipping has an undeniable competitive
advantage in the case of transporting project cargoes, e.g. for replacing plant and
equipment in industrial facilities. Compared to other modes of transport, inland
shipping easily handles large unit weights and larger cargo sizes, therefore cargo
shippers more readily use this mode of transport as long as the minimum navigation
requirements are met. Some cargo shippers from the chemical sector also consider
inland shipping to be the safest form of transport.

Road carriage is selected by cargo shippers to transport small or medium

consignments featuring:

1) greater time sensitivity (e.g. due to the lack of storage space on the
premises or closeness of the overseas markets of the origin or destiny of
the cargo),
2) greater dispersion of deliveries due to multiple premises or the changing
range of the cargo shipper’s business activities,
3) considerable dispersion of customers (in the case of cargo shippers being
intermediaries) and ensuing break-up of cargoes,
4) high diversity due to the degree of unitising in the transport process
(cargoes of the same kind, e.g. fertilisers carried both in bulk and in
unitised forms),
5) high unit value (e.g. technologically advanced cargoes with high unit
prices explicitly gravitate to road transport).

Due to the developed road transport infrastructure, each cargo shipper has a good or
very good access to this mode of transport. Means of road transport are able to meet
the requirements that are impossible for other modes, in particular with regard to
handling cargoes that are temperature sensitive (e.g. when a cargo has to be heated
throughout the time of carriage).

In the case of the smallest consignments, road transport does not experience a
competitive pressure of other modes of transport which are unable to offer
 The Decision Tree Approach for the Choice of Freight Transport Mode:
The Shippers’ Perspective in Terms of Seaport Hinterland Connections
456

competitive price terms for such consignments. Just-in-time deliveries also make it
possible for cargo shippers to obtain savings, as they do not need to maintain
extensive storage capacities. In the case of cargoes stored in bonded areas, which are
problematic in terms of transport organisation (a large share of freight costs in the
value of the cargo itself, atypical dimensions, heavy weight), cargo shippers sell
them to final recipients and require them to organise the transport themselves (due to
the dispersion of consignees and the aforementioned characteristics of the cargoes,
the consignees are inclined to choose road transport). Road transport was also
chosen by cargo suppliers on DDP terms, where cargoes are delivered to the
importers’ premises.

5. Conclusions

Most research studies completed so far in the area of making transport decisions
indicated that the key factors decisive for the mode choice are cost and time of
transport. The existing, more in-depth studies, accounting for cargo shippers’
preferences, took into account only the two modes of transport, road and rail. Our
study carried out with the use of the decision tree methodology, based on the
primary research study on the preferences of cargo shippers located in the seaports
hinterland, took into account three modes of transport, road, rail, and inland
shipping, which had a big impact on transport decisions. The research studies have
shown that the major attributes in selecting transport modes by cargo shippers,
taking into account an access to the three modes of transport to the seaports
hinterland, are single consignment size and time pressure, then owning or having
access to barge terminals by cargo shippers, and the annual volume of cargoes
generated by them.

The study has shown that the greater the frequency of requests to transport large
consignments (above 1500 tonnes), the higher the inclination of cargo shippers to
choose rail haulage, which is additionally enhanced by a greater time pressure for
transport. Analogously, in the case of large single consignments, the smaller the
time pressure, the higher the inclination of cargo shippers to use inland shipping. An
issue of key importance for the final choice of rail transport instead of inland
shipping is having a barge terminal, or an access to it, by the cargo shipper.

In the case of medium single consignments (250–1500 tonnes), also the attribute of
time pressure is of key importance. Also, in the case of this cargo group, the higher
the time pressure, the higher the inclination of the cargo shippers to make a decision
on choosing road rather than rail haulage. Simultaneously, lack of time pressure
inclines cargo shippers to consider rail haulage and inland shipping in their transport
decisions. Another factor affecting the choice is the annual volume of transport.
Smaller annual volumes are more likely to gravitate to inland shipping, due to the
tariff policies provided by rail operators who offer better terms to cargo shippers that
generate a higher annual demand for transport.
 Iz. Kotowska, M. Mańkowska, M. Pluciński

457

In the case of small single consignments (less than 250 tonnes), the studies have
shown that road transport is the only option taken into account in decisions made by
cargo shippers.

The analysis of the transport decision making process has also indicated that in
many cases the choice of a transport mode is affected by individual factors that recur
only in the case of some cargo shippers, i.e. capital ties with a specific port
enterprise (which does not have access to all modes of transport) or a transport
company, the unit value of the cargo, its physical and chemical properties, and the
transport safety level. The two latter aspects are usually a result of a very subjective
evaluation by the cargo shippers of various cargoes showing different physical and
chemical properties. For some, road transport will be safer, as the driver supervises
the cargo at all times (high value cargoes), for others inland shipping is safer, as the
carriage takes place away from residential areas (hazardous cargoes), while some
other shippers think that rail is the safest mode of transport due to the limited
number of transshipment in the transport process, which decreases the risk of cargo
quality deterioration (limited shrinkage). Due to the multitude and subjective
character of the factors indicated above, not all of them could be accounted for in the
developed decision tree, even though they have a significant impact on transport
decisions made by particular cargo shippers.

The research study has also shown that not all attributes included in the study by the
authors were relevant when choosing the transport mode to carry cargoes to/from
seaports. For loads carried over distances greater than 250 km, the transport distance
does not affect the choice of route. Similarly, the type of shipper(manufacturer or
trading company) does not affect the transport route. Trading companies rarely have
access to railway sidings, and their deliveries are more scattered. The use of the
decision tree method made it possible to eliminate the less relevant attributes.

A constraint of the applied decision tree method was a relatively small number of
studied entities, nevertheless it was representative, taking into account the potential
of the hinterland of the seaports covered by the study. The study involved all the
entities that generate more than 100,000 tonnes of cargoes per year, and most of
their cargoes was part of the Polish maritime trading.

References:

Brownlee, J. 2014. Machine learning mastery. Available at: http://machinelearningmastery.

com/discover-feature-engineering-howtoengineer-features-and-how-to-getgood-at-
it.
Chislov, O., Bogachev, V., Zadorozhniy, V., Demchenko, O., Khan, V., Bogachev, T. 2019.
Modeling of the rail freight traffic by the method of economic-geographical
delimitation in the region of the south-easter coast of the Baltic Sea. Transport
Problems, 14(2), 77-87.
Cunningham, W.H. 1981. Freight modal choice and competition in transportation: a critique
and categorization of analysis techniques. Transportation Journal, 21(4), 66-75.
 The Decision Tree Approach for the Choice of Freight Transport Mode:
The Shippers’ Perspective in Terms of Seaport Hinterland Connections
458

Dai, W., Ji, W. 2014. A mapreduce implementation of C4. 5 decision tree algorithm.
International journal of database theory and application, 7(1), 49-60.
De Jong, G., Gunn, H., Walker, W. 2004. National and international freight transport models:
an overview and ideas for future development. Transport Reviews, 24(1), 103-124.
Ferrari, C., Parola, F., Gattorna, E. 2011. Measuring the quality of port hinterland
accessibility: The Ligurian case. Transport Policy, 18(2), 382-391.
Gu, Y., Lam, J.S.L. 2013. Port hinterland intermodal network optimisation for sustainable
development: A case study of China. In: International Forum on Shipping, Ports
and Airports (IFSPA). Hongkong.
Halim, R.A., Kwakkel, J.H., Tavasszy, L.A. 2016. A scenario discovery study of the impact
of uncertainties in the global container transport system on European ports.
Futures, 81, 148-160.
Iannone, F.A. 2012. Model optimizing the port-hinterland logistics of containers: The case of
the Campania region in Southern Italy. Maritime Economics and Logistics, 14(1),
33-72.
Janssens, D., Wets, G., Brijs, T., Vanhoof, K., Arentze, T., Timmermans, H. 2006.
Integrating Bayesian networks and decision trees in a sequential rule-based
transportation model. European Journal of Operational Research, 175(1), 16-34.
Kotowska, I., Mańkowska, M., Pluciński, M. 2018. The Competitiveness of Inland Shipping
in Serving the Hinterland of the Seaports: A Case Study of the Oder Waterway
and the Szczecin-Świnoujście Port Complex. In: Scientific and Technical
Conference Transport Systems Theory and Practice, Springer, Cham, 252-263.
Li, H., Hu, X.M. 2008. Analysis and Comparison between ID3 Algorithm and C4.5
Algorithm in Decision Tree. Water Resources and Power, 26(2), 129-132.
Nam, K., Win, E. 2014. Competitiveness between road and inland water transport: the case
of Myanmar. Transport Problems, 9(4), 49-61.
Notteboom, T., Rodrigue, J.P. 2017. Re-assessing port-hinterland relationships in the context
of global commodity chains. In: Ports, cities, and global supply chains, London:
Routledge, 67-82.
Oral, L.O., Tecim, V. 2013. Using Decision Trees for Estimating Mode Choice of Trips in
Buca-Izmir. ISPRS-International Archives of the Photogrammetry, Remote
Sensing and Spatial Information Sciences 1, 139-145.
Ortuzar, J.D., Willumsen, L.G. 1990. Modelling Transport. Chichester, John and Wiley.
Quinlan, J.R. 1990. Decision trees and decision-making. IEEE Transactions on Systems,
Man, and Cybernetics 20(2), 339-346.
Rasouli, S., Timmermans, H.J. 2014. Using ensembles of decision trees to predict transport
mode choice decisions: Effects on predictive success and uncertainty estimates.
European Journal of Transport and Infrastructure Research, 14(4), 412-424.
Rodrigue, J.P., Notteboom, T. 2010. Comparative North American and European gateway
logistics: the regionalism of freight distribution. Journal of Transport Geography
18(4), 497-507.
Samimi, A., Razi-Ardakani, H., Mohammadian, K.A. 2012. A Decision Tree Approach to
Analyze Freight Mode Choice Decisions. In: Proceedings of the 1st European
Symposium on Quantitative Methods in Transportation Systems. Available at:
https://transp-or.epfl.ch/heart/2012/latsis2012_submission_109.pdf.
Tang, L., Xiong, C., Zhang, L. 2015. Decision tree method for modeling travel mode
switching in a dynamic behavioral process. Transportation Planning and
Technology, 38(8), 833-850.
 Iz. Kotowska, M. Mańkowska, M. Pluciński

459

Thill, J.C., Venkitasubramanian, K. 2015. Multi-layered hinterland classification of Indian

ports of containerized cargoes using GIS visualization and decision tree analysis.
Maritime Economics and Logistics, 17(3), 265-291.
Thore, S., Iannone, F. 2012. The Interport: A Logistics Model and an Application to the
Distribution of Maritime Containers. International Journal of Information Systems
and Supply Chain Management (IJISSCM), 5(4), 23-45.
Tundys, B., Rzeczycki, A., Drobiazgiewicz, J. 2018. Strategic decisions in supply chains.
Kraków, edu-Libri Publishing House. 1-169.
Van den Berg, R., De Langen, P.W. 2011. Hinterland strategies of port authorities: A case
study of the port of Barcelona. Research in Transportation Economics 33(1), 6-14.
Wang, G.W., Zeng, Q., Li, K., Yang, J. 2016. Port connectivity in a logistic network: The
case of Bohai Bay, China. Transportation Research Part E: Logistics and
Transportation Review 95, 341-354.