Application of Graph Theory in an Intelligent Tutoring System for Solving Mathematical Word Problems.pdf

Eurasia Journal of Mathematics, Science & Technology Education, 2016, 12(4), 687-701
Copyright © 2016 by iSER, International Society of Educational Research
ISSN: 1305-8223
Application of Graph Theory
in an Intelligent Tutoring
System for Solving
Mathematical Word Problems
Vasif V. Nabiyev, Ünal Çakiroğlu, Hasan Karal, Ali K. Erümit & Ayça Çebi
Karadeniz Technical University, TURKEY
Received 24 December 2014Revised 27 January 2015 Accepted 13 March 2015
This study is aimed to construct a model to transform word “motion problems” in to an
algorithmic form in order to be processed by an intelligent tutoring system (ITS). First;
categorizing the characteristics of motion problems, second; suggesting a model for the
categories were carried out. In order to solve all categories of the problems, graph
theory including backward and forward chaining techniques of artificial intelligence
were utilized. The study outlines the adoption of graph theory in to the motion problems
and put forth some evidence that the model solves almost all of the motion problems. In
conclusion, the recommended model can be suggested to be used in educational
software in the problem solving context.
Keywords: intelligent tutoring systems, graph theory, mathematics education, motion
problems, problem solving
INTRODUCTION
The basic aim of mathematics education was described as “to bring mathematical
knowledge and skills that are required by daily life to the individual, to teach
students problem solving and to bring them a way of thinking that handles incidents
including a problem-solving approach”. For this reason, problem-solving skills take
an important place among mathematical skills (Baykul, 2004; De Corte, 2004).
Therefore, this issue was carried to the center of the mathematics curriculum at
multiple levels from primary school. Indeed, Nation Council of Teachers of
Mathematics (NCTM) standards also indicate that problem-solving skills have higher
priority in teaching mathematics. Therefore, there is a consensus among
mathematics educators on enhancing the problem solving skills of students being
the primary objective of education (Cai, 2003; Karataş & Güven, 2004). In the mental
process of the problem-solving process; understanding the information presented,
constructing relationships between the information, creative and reflective thinking,
analysis and the use of synthesizing skills are the main sub-processes (Soylu &
Soylu, 2006).
Correspondence: Ünal Çakiroğlu,
Fatih Faculty of Education, Karadeniz Technical University, Söğütlü, Trabzon, Turkey.
Phone: +904628716922
E-mail: cakiroglu@ktu.edu.tr
doi: 10.12973/eurasia.2015.1401a

V. V. Nabiyev
688 © 2016 iSER, Eurasia J. Math. Sci. & Tech. Ed., 12(4), 687-701
For the solution process of problems; Polya
(1957) recommended a framework that contains
the stages of understanding the problem, selecting
a strategy for the solution, the application of the
strategy and the evaluation of the solution. Within
this framework in many studies students were
indicated to encounter a number of difficulties in
understanding the concepts that problems include
and the relationships between the concepts while
they are solving problems (Chiu & Klassen, 2008).
According to Polya, one of the most important
causes of these difficulties is that; students do not
perceive problem solving as a hierarchic process.
This idea was supported by some research studies.
Verschaffel, De Corte, and Viersraete (1999)
determined students’ difficulties with the problem-
solving process while they were solving word
problems. They stated that; traditional arithmetic
word problems cause students to solve word
problems carelessly, superficially and by
conducting routine processes while students
identify the correct arithmetic operations that is
needed to solve an extensive word problem.
In this sense, students often solve word
problems using direct calculation. They do not think over the content of the problem
sentence; because the problems can be solved without deliberating over the
importance of problem content in most textbooks (Greer, 1997; Nancarrow, 2004).
Thus, students do not often find a relationship between school mathematics and
their daily lives. This is because either they have not been taught that there is a
relationship, or because they have not been asked to solve problems which allow
them to use their real-life knowledge while they make decisions about the solution.
If students can be directed to understand the solutions of problems, and to accustom
to use the problem solving steps, they can solve and conceptualize word problems
more easily (De Corte, 2004; Nancarrow, 2004). In this sense Kilpatrick states that;
“a student’s success in problem solving depends on his development of the skills in
problem-solving processes” (Kilpatrick, 1985).
Many efforts were provided for enhancing the problem solving (Cai, 2003; Chiu &
Klassen, 2008; Kılıç & Samancı, 2005). In this sense motion problems were taken
into consideration in problem solving domain of mathematics curriculums.
Mathematics educators think that motion problems are valuable because of
consisting different types of questions, different solution ways and supporting
students’ critical thinking with non-routine features. In addition, mathematics
curriculums suggest that students can construct mathematical knowledge in
accordance of constructivist approach. In this approach in some cases, using
interactive educational software is offered for designing such constructivist learning
environments. Therefore, the teaching motion problems in such a way may allow
students to have experience in problem solving processes by educational software.
In recent years the educational software tended to be designed by using ITS
components. Of course, programmability and the usability of the software are also
associated with the chosen topic. In this regard, motion problems within their
nature serve students to deal with different kinds of problems with different
characteristics. Some efforts were provided for analyzing the problems in order to
carry them in executable form for software. These kinds of software are
recommended to present causal feedback for students to understand and eliminate
State of the literature
 Problem solving is one of the difficult
processes for mathematics instruction,
because; students cannot perform
hierarchically in the process.
 Graph theory is actively used in various areas
biochemistry, engineering and computer
sciences; that give hints about its use in
problem solving.
 Intelligent Tutoring Systems are one of the
favorite tools appearing as software for using
in mathematics problem solving applications.
Contribution of this paper to the literature
 A usable categorization is developed for
motion problems
 An programmable model is developed by
applying graph theory for motion problems
 An artificial intelligence techniques and
graphs are used together for educational
purposes.

Using graph theory in an intelligent tutoring system
© 2016 iSER, Eurasia J. Math. Sci. & Tech. Ed., 12(4), 687-701 689
their mistakes. By this way students can have experience on the new kind of
solutions using feedbacks in different stages of problem solving stages. This study
recommends a model that transforms problems into a form that can be processed by
ITS, with analyzing motion problems. A computer cannot think and make decisions
like humans, especially issues like problem solving is not easy for software. Because
the issue includes, analyzing the problems, working on problem sentences as natural
language, finding different solution ways for the problems and organizing the
computer recourses during these processes. For this purpose, considering the
programmability of the problems, computer resources like memory, disc and data
processing should be well organized. In this context, graph theory was used as a
basic framework in this study. In the history of mathematics, Königsberg bridge
problem which is solved Euler's solution in 1735 is considered to be the starting
point of graph theory. Wate-Mizuno (2014) addressed that the first book about
graph theory is published in 200 years later by D.König in 1936 which is named as
Theorie der endlichen und unendlichen Graphen. Since that time, graph theory is
actively used in various areas such as biochemistry (genomics), electrical and
electronic engineering (communication network and coding theory), computer
sciences (algorithm and computation) and operational research (scheduling).
A graph is a structure defined with the help of edges that expresses a set of points
(nodes) and the relationships between these points G=(Ɣ, E), and it can be expressed
as a set of finite Ɣ and E elements. Here, Ɣ elements are referred to as vertices
(vertex) or nodes. E elements are expressed as edges. Each edge in E adjoins two
different nodes in Ɣ. Nodes are shown by circles and edges are shown by lines
(Dharwadker & Pirzada, 2011). A graph structure with its edges and vertices is
shown on Figure 1.
This study was carried out by focusing on the solution of motion problems Ɣ, E in
graph structure and the relationships between these.
MATERIAL AND METHODS
The study comprises two stages about determining the characteristics of motion
problems and suggesting the model that providing a solution in line with these
characteristics.
The characteristics of the motion problems
In order to determine the characteristics of motion problems, the problems were
compiled from the motion problems from 9th grade mathematics textbooks of
Turkish Ministry of Education. Also questions about motion problems from the
ɣ1
ɣ1
ɣ1
ɣ1
e1 e2
e4
e3
G=(Ɣ,E)
Ɣ={ɣ1,ɣ2,ɣ3,ɣ4}
E={(ɣ1,ɣ2),(ɣ1,ɣ3),(ɣ3,ɣ4),(ɣ1,ɣ4)}
e1=(ɣ1,ɣ2)
e2=(ɣ1,ɣ3)
e3=(ɣ3,ɣ4)
e4=(ɣ1,ɣ4)
E={e1,e2,e3,e4}
Figure 1. An example graph with edges and vertices

V. V. Nabiyev
Entrance Exam to Higher Education to present and motion problems from different
auxiliary sources were chosen. The characteristics were determined by searching
and classifying of 476 motion problem questions. Before the final form of the
classification was given, the above-stated resources were reviewed by five field
experts.
In the process of classification of motion problems, problems were seen to
diversify according to the peculiarities of the number of movements (single vehicle,
two vehicles, etc.), the direction of movement (the same direction, the opposite
direction) and the time of motion (at the same time, different time, and so on). The
basic element in these parameters was determined as the number of vehicles. In
fact, the number of vehicles has a direct influence on the path of the solution to the
problem and different status may arise according to whether the number of vehicles
is one or two. The variables used for different status related to the vehicle numbers
are shown in Table 1.
Motion problems with single vehicle
Two different statuses can be encountered in motion problems with a single
vehicle.
Single Status: Status of a vehicle taking a particular path at a particular
time.
Example 1: How many kilometers does a vehicle travel in 5 hours at 60
km per hour?
Solution 1: It can be found by following the path A1D1M1.
The problem can be solved by being modeled as X=V*T in M1 step.
Two Statuses: Repetition of the situation of a vehicle to take a particular
path at a particular time.
Example 2: A vehicle takes 5 hours to return using the same path that it
had taken 3 hours to drive when traveling at 50 km per hour. So, at what
velocity was it traveling when returned on the same path?
Solution 2: It can be found by following the path A2D2M2.
The problem can be solved by being modeled as XF=X1-X2 = (V1*T1)-
(V2*T2) in M2 step.
Table 1. The variables used in the model
X Path
V Velocity
T Time
XT The total path that the vehicles moved
XG The total of the path on which vehicles moved
XK The remaining path between the vehicles
X1 The path which first vehicle moved X2 The path which second vehicle moved
V1 Velocity of first vehicle V2 Speed of second vehicle
T1 Time of the first vehicle moved T2 Time of the second vehicle moved
X11 Path which first vehicle moved in first status X21 Path which second vehicle moved in first
status
V11 Velocity of the first vehicle in first status V21 Velocity of the first vehicle in second status
T11 Time which first vehicle moved in first status T21 Time which second vehicle moved in first
status
X12 Path which first vehicle moved in second status X22 Path which second vehicle moved in second
status
V12 Velocity of the first vehicle in second status V22 Velocity of the second vehicle in second
status
T12 Time which first vehicle moved in second status T22 Time which second vehicle moved in second
status

Motion problems with two vehicles
There are at least two statuses because of the number of movements in the
solution of motion problems with two vehicles. These two situations turn into 8
models by ramifying with the parameters of the direction, the time and the starting
point. However, with the help of the analysis of these models, two models were
basically seen to emerge. In these models, the parameter of direction determines the
model that will be chosen and the time parameter determines the rule base of the
model that will be used in the solution of the problem. The starting point parameter
allows determining whether the intermediate variable X, which indicates the
distance between the vehicles in the questions regarding the vehicles catching each
other, will take place in the model. Motion problems were investigated on the
perspective of these parameters, a scheme occurs as outlined in Figure 2.
Thus, considering the above-stated parameters, the path to be followed for a
sample question with two vehicles is as follows:
Example 3: The distance between cities A and B is 40 km. A pedestrian
traveling at a velocity of 5 km per hour departs from city A towards city
B and at the same time a cyclist traveling at a velocity of 15 km per hour
departs from city B towards city A. After how many kilometers does the
pedestrian meet the cyclist?
The solution can be found by following the path
A2D3Y2Z3B6M8
The problem can be solved by being modeled as XT= (V1*T1) + (V2*T2)
in M8 step.
If motion problems and parameters were represented in different ways, the
functions that were used to resolve motion problems are basically structured in
three ways. These functions can be expressed as;
1. X=V*T for the structures with single vehicle and single situation, which is the
fundamental solution to motion problems,
2. XF= (V1*T1) – (V2*T2) for the structures with two vehicles and the same
directions, and with single vehicle and two situations,
3. XT= (V1*T1) + (V2*T2) for the questions with two vehicles and opposite
directions.
Considering the solution categories of motion problems, it is possible to divide
motion problems into sub-problems categories using the parameters in each
category. The sample problems for each category were outline in Table 2. The
Figure 2.Model for motion problems with two vehicles

V. V. Nabiyev
solution model was developed according to the variables on the problems such as
number of vehicles, number of status, the directions of the vehicles, time of the
motion and the start point of the vehicles. The three main categories of the problems
and the examples for the categories were shown on Table 2.
After identifying the characteristics of motion problems, a common solution
model was provided for each problem category. A graph structure expressing the
Table 2. Types of motion problems and applied models for solutions
Type Analyses
of the
number
of
questions
Method of
Solution
Model Number
of
Vehicle
s
Number
of
Status
Direction Time/
Start
Point
Example
1.
Type
86 X=V*T A1D1M1 1 1 - - -
How many kilometers will a vehicle
travel in 5 hours at 60 km per hour?
2.
Type
180
XT= (V1*T1) +
(V2*T2)
A2D3Y2Z3B5
M7
2
2
D S S
Two cyclists whose velocities are 8 km/h
different left from the same place at the
same time and traveled in opposite
directions. After an hour from the arrival,
the distance between them is km. What is
the velocity of slower cyclist?
A2D3Y2Z3B6
M8
2 D S D
The distance between cities A and B is 40
km. A pedestrian traveling at 5 km/h, left
from city A and a cyclist traveling at 15
km/h left from city B at the same time
and they were traveling towards each
other. After how many kilometers will
the pedestrian meet the cyclist?
A2D3Y2Z4B7
M9
2 D D S
A vehicle left city A traveling at 60 km/h.
After two hours, if another vehicle leaves
city A in the opposite direction at 70
km/h, what is the distance between them
after 5 hours?
A2D3Y2Z4B8
M10
2 D D D
The distance between A and B cities is
600 km. A vehicle left city A traveling at
50 km/h. Another vehicle left city B
traveling at 75 km/h. They were
traveling towards each other. At what
time will they pass each other?
3.
Type
210
XF= (V1*T1) -
(V2*T2)
A2D3Y1Z1B1
M3
2
2
S S S
Two cars traveled from city A to city B.
They left at the same time. The first car
was traveling at a speed of 30 km/h and
the second car was traveling at a speed of
40 km/h. If the second car arrives at city
B two hours early, what is the distance
between the two cities?
A2D3Y1Z1B2
M4
2 S S D
The distance between cities A and B is
120 km. Two vehicles left cities A and B
at the same time towards each other
traveling at speeds of 85/h and 70 km/h
respectively. How long will it take the
second vehicle to catch the first vehicle?
A2D3Y1Z2B3
M5
2 S D S
A vehicle left city A at a speed of 50
km/h. If another vehicle left city A at a
speed of 80 km/h in the same direction
after 3 hours, how long will it take the
second vehicle to catch the first vehicle?
A2D3Y1Z2B4
M6
2 S D D
The distance between cities A and B is 50
km. A vehicle left city A at a speed of 70
km/h. If another vehicle left city B 3
hours later traveling in the same
direction at a speed of 90 km/h traveling,
how long will it take the second vehicle
to catch the first vehicle?
A1D2M2 1 - - -
A vehicle took 5 hours to return using the
same path that it had traveled on for 3
hours at 50 km/h. How fast did the
vehicle travel on the return journey using
the same path?
D: Different; S: Same

steps and parameters for the solution as a network was used in the model.
Generating the model for the solution of problems
By examining the graphs derived from the literature, the aim of this section is to
express a common drawing type which can provide solutions to different categories
of motion problems. The solutions include parameters such as the number of
vehicles, the number of statuses, the direction, time and the starting point of
movement and various combinations of these parameters. Following basic
principles for the graph structure was provided. These features were identified at
the beginning of the study in order to provide a standard for graphs:
 The direction of the movement must be specified while moving on the
graph.
 The direction of the vehicles (the same direction, the opposite direction,
circular) in the problem can be specified on the edges. For example, if it is
the same direction, (+) or the opposite direction (-) can be put on the
connection edge.
 If a transaction is needed with a constant value on one of the variable
nodes (X, V, T), a node to show it must be generated.
 There may be sub-questions in every node even if the basic nodes are X,
V, T. For example, possible sub-questions for X (showing the path) nodes
may be in the forms of the average path, the path between vehicles or the
total path that vehicles traveled.
 The slots (frame nodes) including sub-questions, information about
nodes and transactions must be created. These slots need to be created
separately for each variable (X, V, T) or even for each node, if it is needed.
The development stage of the model is outlined below:
Preliminary model
The first graph models which were created according to the characteristics of
motion problems and the basic features that are based on X=V*T.
In the first category; the traveled path was expressed as routing nodes which
indicate the parameters of the velocity and time, with arrow directions indicating
the transaction node. The operator in the transaction node enables us to obtain the
result. In the second category; to find the value of the total traveled path, processing
(stated in the transaction node) firstly the values of the velocity and the time for the
first vehicle by arrow directions, the value of the first vehicle traveled is found.
Later, the same transaction is applied for the second vehicle. Finally, the result is
found by processing the values the vehicles traveled in the transaction that is stated
by the arrow directions in the transaction node. Both Type2 and Type3 graph
structures are shown in Figure 3.
XT = (V1 * T1) + (V2 * T2)
V1 T1
X1
V2 T2
X2
XT
*
*
+ or -
Figure 3. Graph structure for Type 2-3

V. V. Nabiyev
In the Type 3 questions, to find the difference between the values of the path that
the vehicles traveled, processing firstly by arrow directions the values of the velocity
and time into the transaction that is stated in the transaction node for the first
vehicle, the value of the first vehicle traveled is found. Then, the same transaction is
processed for the second vehicle. Finally, the result is found by processing the values
of vehicles traveled in the transaction that is stated by arrow directions in the
transaction node.
Although the graph model that was created to express a certain number of
related problems was successful, there were also some limitations as follows: Being
unable to make exactly a common graph figure even for the same categories of
problems, the need for a separate model creation for each categories of problems
and that the way of reading the graph could not fully reflect the questions (having
difficulty in the redaction of the model), lack of flexibility of use, lack of adaption for
all categories of problems and having limitations for possible add-ons in different
types of questions. Therefore, an improved structure to provide a standard for the
graph drawings in one model was needed.
Enhanced model
To eliminate the limitations in the preliminary model, a common model was
developed with a structure that would be able to use Forward Chaining and
Backward Chaining, both of which have an important place in the application of
artificial intelligence. While forward chaining is the application of an inductive
approach towards conclusions from the data, backward chaining is the application
of a deductive approach directly for the conclusion. After the experiences in using
the preliminary model, a three-node graph which includes almost all types of motion
problems and their combined adaptations was developed in the enhanced model.
The basic graph structure of Enhanced model was illustrated in the circles on Figure
4. Forward chaining and backward chaining graph models for the solution of an
Example 4 was shown on Figure 4.
Figure 4. Graph model for problem solving with forward and backward chaining

Example 4: Two vehicles, whose velocity difference is 30 km/h, leave
from points A and B traveling towards each other. After their departure,
they cross at point C after 6 hours. The faster vehicle arrives at point B 4
hours after this encounter. What is the distance between A and B?
As the backward chaining technique follows the sub-nodes which are only
necessary in order to find the result; starting from the desired final result node it
scans the required nodes for final result. This kind of travelling among the nodes is
quite fast and provides reaching the result directly. However, even if this situation
provides solution quickly in designing software, it will lead to different possible
solutions being neglected and even cause the system to enter into an infinite loop for
the problems whose solutions cannot be found directly.
The forward chaining method tries to reach the final result by determining the
blank sub-nodes and starting respectively from these blank nodes repeatedly. As a
result, the system starts again for each blank sub-node, and provides the system to
be operated more slowly while acting in software. However, this disadvantage
actually enables the system to find all the more stable and different possible
solutions. In the enhanced model, by conducting forward and backward chaining
together, possible solution ways can be reached.
Final model
In the final model, variables were presented on different nodes according to their
values and the problem was illustrated in graph components. Therefore, a table of
the representations was prepared after the developing the basic graph structure. In
accordance with the general graph features, the representations and meanings of
the structure were shown in Table 3.
Graph structure is formed in two-vehicle cases in motion problems at the same
time in opposite direction use of the basic graph structure was shown on Figure 5.
The proposed model was tested on various categories of motion problems. Some
adjustments were needed to maximize the solution power of the model. However,
Table 3. The meanings of the nodes relative to the used graph
Node X
if value of the distance is known
if value of the distance is not known
if value of the distance is parametric
Node V
if value of the velocity is known
if value of the velocity is not known
if value of the velocity is parametric
Node T
if value of the time is known
if value of the time is not known
if value of the time is parametric
Process Node Smaller than other nodes and in orange color

V. V. Nabiyev
these adjustments revised with the basic graph model, some other supplements
were still needed. In particular, some supplements were made to the basic graph
structure in parametric problems (the values of variables were not directly
numerical but symbolized by letters) and where vehicles were evaluated relative to
each other. The supplementary graph is needed to find the parameter in the
parametric problems, which may be called as “equivalence graph” and after
inserting the equivalence graph to the general graph, the resulting structure was
shown in Figure 6.
In this structure, as a result of parametric values from below, for example when
there is a need to find X1 and X2, the parametrical expressions V1*T1 and V2*T2 are
directly equalized to XT, and X1 and X2 are subtracted. Due to the subtraction, we
called this operation as “absorption rule”. Process operators coming from below are
connected to the transaction node going to XT by skipping X1 and X2. The
absorption rule is used in the structure created by adding the equivalence graph to
the general graph structure.
Though some requirements were provided by supplement graphs for different
categories of problems, the fact that the basic graph was not changed. The
experimental results on the final graph put forth some evidences about the stability
of the model. In light of the examples given, the different operation steps and 476
different motion problems and with the experiments that have been carried out by
Figure 5. Graph model for two vehicles in opposite direction at time same time
Figure 6. Graph model for two vehicles in opposite direction at time same time

three different experts, it is concluded that the model provides successful and valid
for the solutions for the different categories of motion problems. Two different
examples are provided in example 5 and example 6 and depicted in Figure 7.
An example problem about using XT and XK were inserted to general graph and
form the new structure was shown on Figure 8. It involves problems where the path
between vehicles was needed accept the equivalence graph, with the opposite
direction type motion.
Example 5: The distance between A and B cities is 720 km. Two vehicles,
with velocities of 70 km/h and 50 km/h, leave at the same time and in
opposite directions from these cities. Find the distance between the
vehicles after 5 hours?
Example 6: Part of a 500 km route is asphalt road and part is less quality
road. A vehicle completes the route in 8 hours at 40 km/h on the less
quality road and at 70 km/h on the asphalt road. Find the length of the
asphalt route?
The final graph model applied for Example 5 and Example 6 was shown on
Figure 8.
The solution for the problem in Example 5 is XT= (V1*T1) + (V2*T2). By
following the transactions of number 1234, firstly the path that the first
vehicle traveled (X1) and then the path that the second vehicle traveled (X2) is
For two vehicles
(Example 5)
For one vehicle
(Example 6)
Figure 7. Move in opposite directions at the same time and single vehicle movement
Figure 8. Graph model with forward chaining (Final model) of Example 5 and Example 6

V. V. Nabiyev
found. After finding the sum of these, the total path that the vehicles traveled (XT) is
found. It is subtracted from the total path and the path between the vehicles (XK) is
found.
The solution for the problem in Example 6 is X= V*T is the solution to this
problem. First of all, T1 and T2 values must be parametrically determined in the
graph structure. By following the transactions of number 123, firstly the path
that the first vehicle traveled (X1) and then the path that the second vehicle traveled
(X2) is parametrically found. After equalizing these parametric values to each other,
first the parameter variable (T value) and then the length of the asphalted part of
the path (X2) is found.
DISCUSSION AND CONCLUSION
In this study, the suggested model can be thought as a new application area of
graph theory. In this sense; graph theory applications in mathematics, science and
other fields of technology are becoming increasingly important. It is used to solve
many complex problems in biochemistry (Butenko & Wilhelm, 2006; Bonnici et al.,
2013; Lancia et al., 2001), computer science (Shirinivas, Vetrivel, & Elango, 2010),
GSM technology (Pirzada & Dharwadker, 2007), medical education (Giani et al.,
2007), mining (Ye et al., 2012), architectural design (Poyias & Tuosto, 2015; Roth &
Hashimshony, 1988; Zboinska, 2015), geologic models (Hirsch & Schuette, 1999;
Phillips, Schwanghart, & Heckmann, 2015), urban planning (Almeida, Morley, &
Dowman, 2007; Foltête, Girardet, & Clauzel, 2014), banking (Lin, Tzeng, & Chin,
2011) and simulation of electrical systems (Šarga et al., 2012; Benchouia et al.,
2014). In sum, it is seen that graph theory is being used for the solution of many
complex and extensive problems in daily life. However, a limited number of studies
based on graph theory can be seen in education. In this context; graphs have been
used for modeling students’ using on the web for educational purposes (Gwiszdka &
Spence, 2007; Juvina & Herder, 2005; Juvina & van Oostendorp, 2006; Somyürek,
Güyer, & Atasoy, 2008). In such a research study, Sümersan Seyhanlı (2007) has
focused on the use of the graph theory by the teacher in teaching probability subject.
In this study, the teacher has benefited from the visual representation of probability
problems (representation of all the possibilities in the tree structure) in order that
students could better understand the problems. In this way, students have been able
to find easily the solution identifying the known and the unknown data in the
problems. Students who worked the subject of probability with the graph theory
have outperformed than those who processed traditionally.
This study suggested a model for solution of motion problems with ITS. Two
important conclusions can be asserted for this model. First; this model can be
applied on almost all categories of motion problems and second; this model can
facilitate the programming of motion problems in an ITS. By providing an ITS, the
suggested model can detect the difficulties in the process of solving motion
problems can be detected in the steps. By detailed analysis the solutions for motion
problems can be based on suggested model based on graph theory. Each variable in
motion problems was considered as a node evaluating the graphs as basic structures
defined with the help of edges. The model expresses a set of points (nodes) and the
relationship between these points (Nabiyev, 2012) producing solutions on the basis
of relationships between these nodes.
In this study, artificial intelligence techniques and the graph theory together have
been used to solve the problem beyond the understanding of the problem. Unknown
nodes have been provided to be found by the help of known nodes (variables in the
problems) in the process of the solution of the problem in generated graph
structures. All possible solution ways of motion problems can be modeled with the
graph structures. Therefore, the model can support students to achieve easily

complex cases in the process of problem solving. In this process, the students can be
directed towards their solutions by determining their mistakes by the software.
An important contribution of this study is that; it suggests a structure where
word problems are to be modeled in software. The graph structure which has been
developed for the solution of motion problems in the scope of the study can produce
solutions for problems including variables of the number of vehicles, the number of
situations, the direction, the time and the starting point of moving, and several other
variations of these variables. However, there is another important question that
needs to be asked in this process. It is the question “What is the programmability of
the graph structure?” This question gives programmers an idea related to the
productivity of the graph structure. In this sense, programmers usually use artificial
intelligence techniques developed for intelligent systems. The optimal algorithm for
the solution should be preferred during programming. This is because it is one of the
most fundamental realities of the software world that a problem can be solved
through various algorithms and also similar problems can be solved by different
algorithms. The important point here in this regard is that; a prospective program
produces the most productive result using the least of the computer’s resources.
This can be shortly summarized as the minimum level of memory, using low disc
space and producing solution at maximum speed. The provided graph model may be
shown as one of the optimal solution methods that will be chosen for the software
about the solution of motion problems. Furthermore, being found as the most
suitable solution for the question with forward and backward chaining algorithms in
the graph structure indicates the programmability of this model.
For future work; an ITS on the basis of suggested model will be constructed for
students’ use in problem solving context. Therefore, ITS in the future will contribute
to eliminate the difficulties experienced by students in problem solving process.
Thus, this study can support on the efforts about the aim of “developing problem-
solving skills”, which is one of the general aims of teaching mathematics.
ACKNOWLEDGEMENTS
This research is supported by The Scientific and Technological Research Council
of Turkey (TÜBİTAK).
REFERENCES
Almeida, J. P., Morley J. G., & Dowman, I. J. (2007). Graph theory in higher order topological
analysis of urban scenes. Computers, Environment and Urban Systems, 31(4), 426-440.
doi:10.1016/j.compenvurbsys.2006.03.005.
Baykul, Y. (2004). Teaching mathematics in pimary education. Ankara: Pegem A Publishing.
ISBN 978-605-364-342-5.
Benchouia, N. E., Elias, H. A., K Hochemane, L., & Mahmah, B. (2014). Bond graph modeling
approach development for fuel cell PEMFC systems. International Journal of Hydrogen
Energy, 39, 15224-15231. doi: 10.1016/j.ijhydene.2 014.05.034.
Bonnici, V., Giugno, R., Pulvirenti, A., Shasha, D., & Ferro, A. (2013). A subgraph isomorphism
algorithm and its application to biochemical data. BMC Bioinformatics. 14(7), 1-13
Butenko, S., & Wilhelm, W. E. (2006). Clique-detection models in computational biochemistry
and genomics. European Journal of Operational Research, 173, 1-17. doi:
10.1016/j.ejor.2005.05.026
Cai, J. (2003). Singaporean students mathematical thinking in problem solving and problem
posing: An exploratory study. International Journal of Mathematical Education in
Science and Technology, 34(5), 719-737. doi: 10.1080/00207390310001595401.
Chiu, M. M., & Klassen, R.M. (2008). Relations of mathematics self-concept and its calibration
with mathematics achievement: Cultural differences among fifteenyear- olds in 34

V. V. Nabiyev
countries. Learning and Instruction, 20(1), 2-17.
doi:10.1016/j.learninstruc.2008.11.002.
De Corte, E. (2004). Mainstreams and perspectives in research on learning (mathematics)
from instruction. Applied Psychology, 2(53), 279-310. doi: 10.1111/j.1464-
0597.2004.00172.x.
Dharwadker, A., & Pirzada, S. (2011). Graph theory. Amazon. ISBN 1466254998.
Foltête, J. C., Girardet, X., & Clauzel, C. (2014). A methological framework fort he use of
landscape graphs in land-use planning. Landscape and Urban Planning, 124, 140-150.
doi: 10.1016/j.landurbplan.2013.12.012.
Giani, U., Brascio, G., Bruzzese, D., Garzillo, C., & Vigilante, S. (2007). Emotional and cognitive
information processing in web-based medical education. Journal of Biomedical
Informatics, 40(3), 332-342. doi:10.1016/j.jbi.2006.11.004.
Greer, B. (1997). Modelling reality ın mathematics classrooms: the case of word problems.
Learning and Instruction, 7(4), 293–307. doi: 10.1016/S0959-4752(97)00006-6.
Gwizdka, J., & Spence, I. (2007). Implicit measures of lostness and success in web navigation.
In Interacting with Computers, 19(3), 357-369. doi: 10.1016/j.intcom.2007.01.001.
Juvina, I., & Herder, E. (2005). The impact of link suggestions on user navigation and user
perception. In UM2005 User Modeling: Proceedings of the Tenth International
Conference, 483-492. doi: 10.1007/11527886_66.
Juvina, I., & van Oostendorp, H. (2006). Individual differences and behavioral metric
involved in modeling web navigation. Universal Access in Information Society, 4(3), 258-
269. doi: 10.1007/s10209-005-0007-7.
Hirsch, L. M., & Schuette, J.F. (1999). Graph theory applications to continuity and ranking in
geologic models. Computers & Geosciences, 25(2), 127-139. doi: S0098-3004(98)00116-
2.
Karataş, İ., & Güven, B. (2004). Determination of 8th students’ problem soluting skills: A case
study. Journal of National Education, 163.
Kılıç, D., & Samancı, O. (2005). The usage of problem solving method in social knowledge
lesson given in primary schools. Journal of Kâzım Karabekir Education Faculty, 11, 100–
112.
Kilpatrick, J. (1985). A restrospective account of the past 25 years of research on teaching
mathematical problem solving. In E.A. Silver (Ed.), Teaching and learning mathematical
problem solving: Multiple research perpectives. Hillsdale, NJ: Lawrence Erlbaum
Associates.
Lancia, G., Bafna, V., Istrail, S., Lippert, L., & Schwartz, R. (2001). SNPs Problems, Complexity
and Algorithms. 9th Annual European Symposium (ESA 2001), 2161, 182-193. Berlin
Heidelberg: Springer-Verlag.
Lin C. S., Tzeng, G. H., & Chin, Y. C. (2011). Combined rough set theory and flow network
graph to predict customer churn in credit card accounts. Expert Systems with
Applications, 38(1), 8-15. doi:10.1016/j.eswa.2010.05.039.
Nabiyev, V. V. (2012). Artificial intelligence. Ankara: Seçkin Publishing. ISBN 987-975-02-
2034-0
Nancarrow, M. (2004). Exploration of metacognition and non-routine problem based
mathematics instruction on undergraduate student problem solving success.
(Unpublished doctoral thesis). The Florida State University, Florida.
Pirzada, S., & Dharwadker A. (2007). Applications of graph theory. Journal of The Korean
Society for Industrial and Applied Mathematics. 11(4), 19-38. doi: 10.12941/jksiam
Polya, G. (1957). How to solve it? (2 nd ed.). Princeton, N.J.: Princeton University Press.
Roth, J., & Hashimshony, R. (1988). Algorithms in graph theory and their use for solving
problems in architectural design. Computer-Aided Design, 20(7), 373-381.
doi:10.1016/0010-4485(88)90214-X.
Somyürek, S., Güyer, T., & Atasoy, B. (2008). The effects of ındividual differences on learner’s
navigation in a courseware. Turkish Online Journal of Educational Technology, 7(2), 32-
40.
Soylu, Y., & Soylu, C. (2006). The role of problem solving in mathematics lessons for success.
Inonu Unıversıty Journal of The Faculty of Education, 7(11), 97–111.
Šarga, P., Hroncová, D., Čurilla, M., & Gmiterko, A. (2012). Simulation of Electrical System
using Bond Graphs and MATLAB/Simulink. Procedia Engineering, 48, 656-664.
doi:10.1016/j.proeng.2012.09.567.

Shirinivas, S. G., Vetrivel, S., & Elango, N. M. (2010). Application of graph theory in computer
science an overview. International Journal of Engineering Science and Technology, 2(9),
4610-4621. ISSN 0975-5462.
Sümersan Seyhanlı, S. (2007). The effect on the student achievement of graph theory in
teaching of the unit of “probability” of primary school 8th. (Master thesis). The Science
Learnings Institute, Balıkesir.
Phillips, J. D., Schwanghart, W., & Heckmann, T. (2015). Graph theory in the geosciences.
Eart-Science Reviews, 143, 147-160. doi: 10.1016/j.earscirev.2015.02.002.
Poyias, K., & Tuosto, E. (2015). A design-by-contract approach to recover the architectural
style from run-time misbehaviour. Science of Computer Programming, 100, 2-27. doi:
10.1016/j.scico.2014.10.005.
Verschaffel, L., De Corte, E., & Viersraete, H. (1999). Upper elementary school pupils’
difficulties in modeling and solving nonstandard additive word problems involving
numbers. Journal for Research in Mathematics Education, 3(30), 265-285. doi:
10.2307/749836.
Wate-Mizuno, M. (2014). Mathematical recreations of Dénes König and his work on graph
theory. Historia Mathematica, 41, 377-399. doi: 10.1016/j.hm.2014.06.001.
Ye, Q., Wang, H., Jia, Z., & Pi, Y. (2012). Graph theory and its application in optimization of gas
drainage system in coal mine. Procedia Engineering, 45, 339-344. doi:
10.1016/j.proeng.2012.08.168.
Zboinska, M. A. (2015). Hybrid CAD/E platform supporting exploratory architectural design,
Computer-Aided Design. 59, 64-84. doi: 0.1016/j.cad.2014.08.029.


Application of Graph Theory in an Intelligent Tutoring System for Solving Mathematical Word Problems.pdf

More Related Content

Similar to Application of Graph Theory in an Intelligent Tutoring System for Solving Mathematical Word Problems.pdf (17)

More from Valerie Felton (20)

Recently uploaded (20)

Application of Graph Theory in an Intelligent Tutoring System for Solving Mathematical Word Problems.pdf