Contemporary Problems of Social Work


Adaptive models of computer training

Автор/Author: Gdansky N. I., Altimentova D. Y.

Скачать/Download | Посмотреть онлайн/View online

Список литературы/References:

1. Avanesov B. C. the Composition of test tasks: a textbook for teachers of universities, colleges,

schools, teachers, schools, gymnasiums and lyceums. - M.: testing Center, 2002. - P. 238

2. 2. Anastasi A. Psychological testing / A. Anastasi and S. Urbina. - SPb.: Peter, 2003.

3. 3. Bespalko V. P. Education and training with the participation of computers (the pedagogy

for the third Millennium) / V. P. Bespalko. - M.: Publishing house of the Moscow psihologosocial

Institute; Voronezh: MODEK, 2002.

4. 4. Rabinow, E. H. personalized Adaptive system training for engineering students: monograph.

- M.: Mechanical engineering, 2009. - P. 258

5. 5. Sviridov A. P. Standardized methods on the example of monitoring and diagnosing

knowledge: monograph. – M.: Publishing house of Rostov state construction University,

2011. – P. 294

6. Kryukova E. M. Optimization of the management of the Corporation using web technologies.//

Service plus. 2011. No. 4. pp. 102-108.

7. Kryukova E. M., G. V. Krasilnikova areas of internal corporate use of blogging in the activities

of the companies.//Service plus. 2011. No. 2. pp. 96-104.

8. Kryukova E. M., Krasil’nikova G. V., Kulakova T. V. Social media: the real and potential sources

of threats to economic security of the enterprise.//Bulletin of the Academy of economic

security Ministry of internal Affairs of Russia. 2011. No. 3. pp. 53-57.

9. Maloletko A. N., Kaurova O. V., Kryukova E. M. Yukhin K. E. The role and importance of modern

media in marketing communications.//Proceedings of the Russian state social University.

2014. No. 3 (125). pp. 96-103.

Refernces in Roman script:

1. Avanesov B.C. Kompozicija testovyh zadanij: uchebnik dlja prepodavatelej vuzov, tehnikumov,

uchilishh, uchitelej shkol, gimnazij i liceev. - M.: Centr testirovanija, 2002. - 238 p.

2. Anastazi A. Psihologicheskoe testirovanie / A. Anastazi, S. Urbina. - SPb.: Piter, 2003.

3. Bespal’ko V.P. Obrazovanie i obuchenie s uchastiem komp’juterov (pedagogika tret’ego

tysjacheletija) / V.P. Bespal’ko. - M.: Izd-vo Moskovskogo psihologo-social’nogo instituta;

Voronezh: MODJeK, 2002.

4. Rjabinova, E.H. Adaptivnaja sistema personificirovannoj professional’noj podgotovki studentov

tehnicheskih vuzov: monografija. - M.: Mashinostroenie, 2009. - 258 p.

5. Sviridov A.P. Standartizirovannye metody na primere kontrolja i diagnostirovanija znanij:

monografija. – M.: Izdatel’stvo RGSU, 2011. – 294 p.

6. Krjukova E.M. Optimizacija upravlenija korporaciej s ispol’zovaniem veb-tehnologij.//Servis

plus. 2011. № 4. S. 102-108.

7. Krjukova E.M., Krasil’nikova G.V. Napravlenija ispol’zovanija vnutrennego korporativnogo

blogginga v dejatel’nosti kompanij.//Servis plus. 2011. № 2. S. 96-104.

8. Krjukova E.M., Krasil’nikova G.V., Kulakova T.V. Social’nye media: real’nye i potencial’nye

istochniki ugroz jekonomicheskoj bezopasnosti predprijatija.//Vestnik Akademii

jekonomicheskoj bezopasnosti MVD Rossii. 2011. № 3. S. 53-57.

10. Maloletko A.N., Kaurova O.V., Krjukova E.M., Juhin K.E. Rol’ i znachenie sovremennyh

mediasredstv v marketingovyh kommunikacijah.//Uchenye zapiski Rossijskogo

gosudarstvennogo social’nogo universiteta. 2014. № 3 (125). S. 96-103.

Содержание статьи/Article:

The introduction of new developments in the field of e-learning is one of the main

indicators to assess the success of the institution in modern conditions. Education with using

a computer technology has several features that significantly differ it from the traditional way.

The main drawback is the lack of computer training “real” contact of the student with the

teacher that allows you to build a learning process which based on individual characteristics.

However, a modern computer technology, due to a high-capacity of storage devices and highspeed

processing, allow to us much to expand the sphere of the materials which we used in

the educational process, and through the modern methods of analysis dialogue with user, to

a great extend to adapted the transmission of knowledge to a specific student . This creates




74 75


preconditions for the maximum approximation of the general level of the quality of e-learning

to an individual learning at the present stage of the development of computer technologies.

Practical realization of the individual training - is one of the most important conditions for the

professionals’ training.

The process of learning by the technical profile disciplines, especially those which are directly

connected with the information technologies in need of a great handling in compared with a

traditional structure of it. This is due to the high growth of the total volume of information,

increasing its availability, constant modarnization and complications of computer and other

technologies. Practically, the world is changing every day, the information’s volume is increases,

including new technologies that demand an early and successful understanding of modern

students, who sets a goal to be a good specialists on a job market, professionally literate. That

is why one of the main goal of the modern education is the harmony implementation of these

technologies in all spheres of the educational process.

In order to give effect to the requirements of time and teach the excellent specialists in

the labor market who are ready to immediately do a necessary functions at the workplace,

educational institutions, as well as teachers, who teaches to the disciplines, must solve several

interconnected problems such as:

1) to define in a huge, ever-increasing area of information exactly those of its components,

which the student are real needed;

2) to publish a useful information into a specific form of educational materials (lectures,

seminars and laboratory lectures, tests and etc.), which allows to the students to understand it

in real educational process, which is limited of time constraints and the organization’s forms;

3) to include these knowledge into the whole structure of studying directions (or speciality)

in the best way, and in particular the specific discipline;

4) to create a technology for transmission electronic knowledge of students which full shows

all their features;

5) to use the methods of examination and testing of knowledge, as much as it possible to

take into consideration the opportunities of electronic technologies.

The rapid development of computer technology and computer-based training methodology

can allow considerably improve the quality of computer studying, to bring closer it to an

individual way. Adaptation is the most suitable, the universal variant to improve computer


Using computer knowledge test the most convenient form is a single test – there is only one

correct answer. The adaptation of these single test requires the elaborations of a formal set of

features that can be possible effective computer realization/

For automatic construction of adaptive algorithms of examination it is necessary together

with the traditional estimator to identify the control function of at each step which gives the

transition to the next step. It should to choose the type of control function which will be useful

for software realization.

A computer model of a single test should determine a more flexible approach to the

assessment procedure of knowledge in comparison with the classical way “yes - no”. It was

offered to present the general dependence test result (R) from the user’s actions (V) at the time

of a single test with N possible response options, and one correct answer, represented in the

form as:

R = R({V}, N, {P}) = R+({V}, N) ⋅ R-(N, {P}), (1)

where R - the final student’s mark for the students answre, R + ({V}, N) - the basic component

of the mark, the function of estimate, R- (N, {P}) – a fine component of the mark, which reduces

it, {V} – the information, which had been received from students (by one or more variants of the

correct answer), N - number of possible answers, {P} – a fine function.

We will keep in mind as a basic mark “pure” number characteristic of the solution, which

estimates only the truth - close to the correct answer. The classical way is to use in logical


0 if the answer is wrong


1 if the answer is right

⎧ −

= ⎨ − ⎩

This way means that it will be used the real estimate: R∈[0,1]. The final estimate “yes-no” is

actually a logical meaning in the classical version. It is more convenient for the teachers testing,

but at the same time it is much more simplified in comparison with the real estimate and does

not allow subtly evaluate the various nuances of checking knowledge. This disadvantage is

called for eliminating a continuous real estimate.

{V} – the information which was received from the student (a plenty of possible answers).

The classical way is that the plenty {V} must contains one answer.

The expected way is that the student can choose several possible correct answer, {V}, which,

in his opinion, include the correct answer. The transition to multiple answers allows to pay more

attention to the level of knowledge with regard to their temporary forgetfulness, psychological

factors of the examination situation. In this case, the base component of the estimate (mark)

R+ ({V}, N) was offered to make with using the following relation (2). This type of estimate

function R+ ({V}, N) means the estimate (mark), which will be made in case if there is the correct

answer in a group of ⏐V⏐ answers.



0, если V 0 или V {V};


R( V ), V {V}, V 0.

⎪⎧ = ∉ = ⎨ ∈ > ⎪⎩


In general, when you are testing different types of knowledge the fine may be imposed as

on the general violation of testing procedures on the part by student, and the assumption of

semantic wrong answers. When you are testing by personal teacher as a “special testing system”

it has the opportunity to take into account the whole set of the fine factors – if the student has

not done it by himself, has written it of the paper, has used the mobile phone and etc..

Modern computers cannot do such additional control functions. That is why, as a general

violation of testing procedures had been offered to take the time of the answer, because it,

along with the correct answer also allows indirectly assess the level of training of the student.

This additional characteristic response is one of the few that can be effectively tested with

computer testing.

At first it was study linear and quadratic relation of R+ ({V}, N) during the constructing the

evaluation function. Linear has been excluded because it does not give overestimates the R+

at small | V |, for example, for N = 5 and | V | = 2 => R+ = 0.75. The main disadvantage of the

quadratic relation is that even for small values of N it is not monotonically decreasing. For

example, in case of N = 5 at the approach to endpoint function is negative, which contradicts

the definition of the function as a positive assessment. At the point of the N = 4 (for N = 5), the

function prematurely takes zero value, ie, is not monotonically decreasing.

It was offered to use a cubic relation which is for geometric conditions on the curve

R+(1,N)=1; R+(2,N)=0,5; R+(N,N)=0; R+′(N,N)=0 (3)

it gives the next formula:

R+(Х,N)= (N-V)2[(-N 3+9 N 2 -21 N+15)⋅(N-V)+(N2- 6N+7)] / [2(N-1)2(N-2)2] (4)

As the practice shows the optimal value is N = 5. If there is so, on the one hand there is a

small positive estimate at random guessing the answer (0,2). On the other hand, the number of

answers is not too large and does notprolong the process of testing. For more convenient using

76 77


of the value of the function, were made the following discrete values of the function estimation

in the nodes for N = 5:

R+(1,5)=1;R+(2,5)=0.5;R+(3,5)=0,2;R+(4,5)=0,1;R+(5,5)=0. (5)

For the variant of the fine function which was offered it is necessary at first to find the

optimal time of the answer TОПТ(N) in seconds. It was offered to use the next formula for its


TОПТ(N) = ТОТВ + N ⋅ ТВОПР, (6 а)

where ТОТВ – is the time of evaluation on the correct answer on the test’s question , ТВОПР – is

the time of the one of the variant. ТОТВ magnitude and ТВОПР should be assigned based on the

complexity of the issues. For example, if verified review material medium level of complexity, it

is recommended to take:

ТОТВ = 4 ÷ 5, ТВОПР = 1,5 ÷ 2. (6 б)

If you want to use a rather complicated method, or algorithm, then, on the basis of experience,

should be taken:

ТОТВ=10÷20, ТВОПР=3÷5. (6 в)

If you want to construct a fine function at {P} = T it was offered to use the following practical

considerations: 1) for 0 < Т ≤ TОПТ (N) time for a response has not expired, no fine will be imposed

and R-(N, {P}) = 1; 2) when TОПТ (N) ≤ Т ≤ 3TОПТ (N) response time prolonged, indicating a low level

of knowledge or accessing the prompts.

It was suggested:

1) to take the limit waiting time equal to three times the maximum response time;

2) linear decrease of the fine function R-(N, {P}) от 1 ( при T=TОПТ(N)) до 0 (при T=3TОПТ(N));

3) Т > 3TОПТ(N) value R-(N, {P})=0; very exaggerated response time indicates the use of

unacceptable sources to respond to the test. Obtained as a result of the relation has the next


( )





1, если T T ;

R ,{ } 0,5(3-T/T ), если Т Т 3Т ;

0,если Т 3Т .


≤ ⎧⎪

= < ≤ ⎨⎪

⎩ >


You can see the example of the diagrame R-(T), when TОПТ=15sec.

The relation R-(T) for TОПТ=15 sec.

Let’s examine the second level of adaptation for individual control of knowledge, which

which is a flexible construction of the whole process of testing and take a structure of tests

materials, and the correctness of the answers.

Using the principle of adaptation in the traditional teacher knowledge by himself takes the

maximum free time to the teacher and the most exact examination of students’knowledge.

Using the computer takes us an opportunity to increase efficiency of the person who relying

in many times, ie, to eliminate the limitation on the “efficiency” of the teacher. However, the

limited time which we have in the curriculum on knowledge tests and physiological limits,

requires a reasonable limit set by the total number of tests on the same tests. That is why

adaptability in computer testing must take into attention the significant number of opportunity

of asking questions.

Let’s examine the construction of the evaluation and management functions that provide

automatic generation of adaptive algorithms of modular test knowledge, have a fairly simple

structure and convenient for software implementation.

On the other hand is more convenient structure to do information-testing operations

represents a rational the following composition of the basic materials for the semester discipline.

The information block consists of 3 - 5 sections with dimensions from 20 to 40 - 50 pages of

informative text. Each section consists of 3 - 5 subsections, each size from 5 to 12 - 15 pages.

The testing of the teaching material. At a minimum, the test unit studied discipline adopted

module (section), so Testing materials attached to the relevant sub-sections of the studied

subjects. Also provides entrance test to check for vestigial knowledge of the previously studied

subjects such as university and school curricula. In general, the testing unit subsection contains

a set of 10 - 30 questions to test theoretical knowledge and 10 - 30 tasks to test practical skills.

Of the total number of case studies about 20 - 30% identified as tasks of high complication.

Tests are divided into:

1) educational, which conducted independently for the evaluation of knowledge by students,

2) control, conducted under the supervision of a teacher for an objective assessment of the

actual level of knowledge.

Practically all these kinds of tests s are provided using two different algorithms for testing.

Educational Test is always logical to do by modules-sub. The results of the analysis of each

theoretical question and implementation each problem promptly declared to the student. As a

result of the analysis of the all variants are formed trajectory control - select the number and

content of the regular questions and practical tasks. There are a general mark in scores at the

end of test and generall recommendations on re-examination of the material (if it is necessary).

Control test can be done on the sub-sections, sections and throughout the course - depending

on the request formulated by the teacher. In order to reduce the response timeof present

monitoring results are not declared student. At the end of the test each student get a general

mark in scores. It does not matter the level of control system for group analysis of knowledge

because each time formed the partial score of knowledge separately for each sub-section.

The volume of test material is divided into:

1) subsection (module)

2) section (in particular – a final control)

3) all discipline (in particular - the final control).

These possibilities are guarantees by the organization of the information and material

testing for sub-course and taken into account in testing algorithms.

Each unit of the test subsection includes a basic set of 20-30 questions to check up the

theoretical knowledge. If the subsection contains the theoretical material and / or methodology

of solving the problem, the base section of the testing depending on the content of the

educational material should be supplemented with the following materials:

1) with theoretical questions of high level,

78 79


2) with tasks,

3) with tasks of high level.

In such a way each of the two types of testing (training and supervision) should be adaptive

at the same time on two characteristics – on the depth of interrogation – the level of knowledge

test (1-level, 2-level, 3-level) and the degree of coverage of the material (subsection, section,


The adaptation of the first criterion suggested that by using specific individual basic

algorithms that take into account the number of stages of interrogation with a minimum

number of t basic questions in a single step, equal to 2 (t = 2).

The adaptation of the second criterion was offered to realize by a large-scale increasing the

number of tests in the basic version of the algorithm.

For clearness, the basic and derived algorithms testing was offered to represent as trees.

Based on the solving the task, the algorithms which work out by the unit test must be

adaptive to the following set of features:

1) The type of testing,

2) the coverage of the test (it means the volume being checked by general algorithm)

3) the number of types of test materials.

The 1-level test. This test is offered to do with using 1 - 2 phases of the interrogation,

depending on the correctness of the answers on the first base phase. If after the base phase there

is no negative answer, it’s used the second phase of the interrogation with a reduced number

of questions (less than half compared to the base). The second phase of the interrogation is

called specified.

The main tree call the tree, which realizes the algorithm of interrogation with the minimum

number of individual tests. As a main tree of interrogation we will take the structure which

shown on the picture.

The main tree for the 1st module of the test A scaled algorithm with coefficient t=2

Here we can see the minimum possible number of basic questions which is equal 2. The

maximum number of points is 100. The upper row is equally to a base phase of the interrogation,

the lower row is to be confirmed. On the edges indicate the number of correct answers in

the appropriate phase of the interrogation. Next to the nodes are shown a final number of

assessment points.

The tree was made according to the next rules.

1. If all the basic answers are wrong, the final score is 0 points (leftmost vertex in the upper

row), the interrogation stops.

2. If all the basic answers are correct, the final score is 100 balls (rightmost vertex in the

upper row), the interrogation stops.

3. If you receive one right answer in the basic phase (average peak in the upper row), the

mid-term score is 50 balls, the interrogation continues in subdivision phase (lower row).

4. In specifying phase is given by halving the number of questions compared to the baseline


5. In case if it given the wrong answer, the number of points is the same as in the intermediate

phase - 50 (the leftmost vertex in the lower row).

6. When it given a correct answer, the number of points is increased to 70 (as a result of the

percentage of correct answers is 100 * (2/3) ≈ 70) - (rightmost peak in the lower row).

Scaling tree of the interrogation was suggested to do by increasing the number of basic

issues in t times. Setting the parameter t to adapt the test procedure to the scanned volume of

educational material.

It was offered (suggested)

1) when you check up all the discipline to do for the module (sub) the minimum value of t

= 1 or 2,

2) when you check up the section to do, depending on the total number of modules and the

importance of a particular module value t = 2 or 3,

3) when you check up one module (section), you must do the value of t = 3 or 4.

Scaled algorithm with a coefficient t = 2 for modular single-stage test is illustrated in picture

3 on the interrogation’s tree.

When you build an algorithm of 2-level test, you must first include the distribution of the

maximum score obtained in the first and second phase of interrogation for the basic algorithm

of the 2nd level. It was taken a total number of points equal to the first phase 40, the second

- 60. Then the value of the correct answer will be in the first phase is 40/2 = 20 points, in the

second - 60/2 = 30 points. As a fine for mistake in the initial phase of polling in the evaluation

of clarifying questions on the first phase of the suggested value of each correct answer is worth

2 points in the number of times smaller than at the initial interrogation, ie, 20/2 = 10 points. As

a result, an interrogation’s tree of the 2nd-level of the test, according with the aforementioned

conditions it will be have the next view (picture).

The interrogation’s tree of the 2nd test

The scaling of this tree is made the same as the 1-level test.

Similarly, we do the 3rd-level test. It is recommended to use the maximum distribution points

obtained in the first, second and third phases form interrogation: 28 - 32 - 40. Using higher

levels of the test virtually no rationally.

The model of the single alternative test increases the degree of adaptability in comparison

with the classical threshold test due to a more flexible response options assessment and

consideration of timing response. The classical model of computer-based testing involves simple

copying actions of the teacher. In the proposed model for the implementation of adaptability

in the extended version used the possibility of a computer system. This allows you to more fully

realize the personalized computer training in the first stage of examination - at the level of a

single test.

The adaptive algorithms single test and alignment of the testing process can significantly

bring quality computer education to the individual. More objective identification of the current

level of preparation of each student can be used for optimum correction of knowledge, which is

especially important in group forms of education.

Ключевые слова/Tags1: adaptation, testing, computer learning, single test testing algorithm.