Abstract
Content
- Introduction
- 1. Goal and tasks of the research
- 2. Structure of the mental portrait
- 3. Building and researching a neural network model
- 3.1. Designing a Model
- 3.2. Building a Model
- 3.3. Results
- Conclusions
- References
Introduction
The work deals with the issues of forecasting and analysis of the quality of student’s education.
Evaluation of the effectiveness of the processes of preparing specialists is very important for the analysis and management of the university as an educational environment.
This task is difficult to formalize because of the heterogeneity, intellectuality and dynamism of interaction of subjects, so to describe the exact behavior of such a system by classical mathematical methods is very problematic.
1. Goal and tasks of the research
This work is devoted to solving the problem of forecasting the quality of students' professional education depending on their personal characteristics and other factors. It is solved on the basis of the application of artificial neural networks and is reduced to the development of a neural network model, which is able to functionally describe the dependence of the student's professional knowledge and skills on personal factors.
2. Structure of the mental portrait
Among the important factors that influence the learning process of a student are a combination of individual psychophysiological features of a person. To identify them, popular psychological methods of personality analysis were used:
- type of temperament [5];
- type of motivation;
- intelligence quotient [6];
- social intelligence;
- special abilities;
- level of creativity [7].
All 6 tests contained more than 200 tasks, which significantly complicated the presentation of tests and increased the testing time with a large number of interviewed students. In order to speed up the process of personality testing, a system was developed to automate the interviewing of students and the processing of responses to psychological tests [8].
3. Building and researching a neural network model
3.1 Designing a Model
The amount of residual knowledge depends on the student's mentality and some other factors described in the previous section. This connection is difficult to formalize, i.e. it is difficult to describe it mathematically. In such cases, as mentioned above, it is advisable to use a neural network, which will reveal the existing connection. The following objective information is available for teaching the neural network:
- a psychological portrait of a student's mentality;
- educational program;
- knowledge assessment criteria;
- examination results.
The neural network will be trained on the mental portraits of a group of students and their grades.
Research on the quality of neural network forecasting was based on mental portraits that did not take into creativity and some other factors. Input signals are the main mental characteristics of students, which are described in the previous section, and statistics of student attendance. Neural network input signals form vector X = (x1, x2, …, x6), which components are described in Table 1.
Input signal name | Signal code |
---|---|
Type of motivation | x1 |
Type of temperament | x2 |
Intelligence quotient (IQ) | x3 |
Social intelligence | x4 |
Special abilities | x5 |
Student attendance | x6 |
The output neural network generates signals that determine the predicted examination score. In the model, the evaluation is encoded by a four-digit binary code (Table 2).
Examination score | Binary code |
---|---|
2 | 0 0 0 1 |
3 | 0 0 1 0 |
4 | 0 1 0 0 |
5 | 1 0 0 0 |
3.2 Building a Model
The Deep Learning Toolbox for the MATLAB environment [10] was a toolkit for creating a neural network model.
The neural model of examination evaluation forecasting is based on the Feed forward neural network [9]. According to the researches [11], for the function approximation task, it is enough to use 2–3 layers to ensure the implementation of any nonlinear dependence between input and output. In the model described in this article, 2 neuronal layers were used.
3.3 Results
Prediction of students' progress was made on the educational set, consisting of mental portraits and examination scores of 60 students.
The results show that in the array of students with a grade of 3
, due to the large number of training examples, the probability of accurate prediction at 100 restarts for training reaches the value of 0.9, which cannot be seen in other arrays. This experiment emphasizes the need to further seek to increase the amount of training set.
Conclusions
The task of predicting the performance of students by their mental portraits is set.
The prototype of the system of building a mental portrait of a student on the basis of psychophysiological testing of personality is developed.
On the basis of the results of psychometric assessment of students' personalities the educational set for the adjustment of the neuromodel is constructed. The executed researches have shown possibility of application of the neural network approach to the decision of a problem of prognosis of students' progress and a direction of the further work on improvement of quality of model.
When writing this essay, the master's work is not yet complete. Final completion: May 2020. The full text of the work and materials on the topic can be obtained from the author or his manager after the specified date.
References
- Калинин А.М., Тарасов В.Г. Использование данных электронного обучения для оценки и прогнозирования результатов. V Международная научно-практическая конференция
Электронное обучение в непрерывном образовании, ЭОНО–2018
(Россия, Ульяновск, 18–20 апреля 2018 г.): сборник научных трудов. – Ульяновск: УлГТУ, 2018. – С. 493–501. - Федяев О.И. Нейросетевая модель процесса профессионального обучения молодых специалистов // Открытые семантические технологии проектирования интеллектуальных систем (OSTIS–2015): материалы 5-й межд. научно-техн. конф. (Минск, 19–21 февраля 2015 г.). – Минск: БГУИР, 2015.
- Русаков С.В., Русакова О.Л., Посохина К.А. Нейросетевая модель прогнозирования группы риска по успеваемости студентов первого курса // Современные информационные технологии и ИТ–образование, Vol.14, N 4, 2018. – p.815–822.
- Федяев, О.И. Формирование зависимости остаточных знаний студентов от их ментальности с помощью нейронных сетей / Федяев О.И., Грабчук О.П., Елифёров В.В. // Труды конференции ИАИ–2015, КПИ, Киев, 2015. – С. 258–265.
- Айзенк, Г. Как измерить личность: учеб. пособие / Айзенк Г., Вильсон Г. – М.: Изд-во Когнито–Центр, 2000. – 27 с.
- Айзенк Г. Новые тесты IQ – М.: Изд-во
ЭСКМО
, 2003. – 189 с. - Ильин Е.П. Психология творчества, креативности, одарённости. – СПб.: Питер, 2004. – 537 с.
- Гончаров, К.Д. Система построения ментального портрета студента с речевым интерфейсом [Электронный ресурс] / Гончаров К.Д., Федяев О.И. //Информатика, управляющие системы, математическое и компьютерное моделирование в рамках IV форума
Инновационные перспективы Донбасса
(ИУСМКМ – 2018): IX Международная научно-техническая конференция, 22–24 мая 2018, г. Донецк: / Донец. национал. техн. ун-т. – Донецк: ДонНТУ, 2018. – Режим доступа: http://iuskm.donntu.ru . - Deep Learning: Feedforward Neural Network [Электронный ресурс] – Режим доступа: https://towardsdatascience.com
- Deep Learning in MATLAB [Электронный ресурс] – Режим доступа:https://www.mathworks.com
- Saurabh Karsoliya Approximating Number of Hidden layer neurons in Multiple Hidden Layer BPNN Architecture. // International Journal of Engineering Trends and Technology, Vol.3, N 6, 2012. – p.714–717.
- Бэстенс Д. Э., Ванн ден Берг В. М., Вуд Д. Нейронные сети и финансовые рынки: принятие решений в торговых операциях. – М.: ТВП, 1997. – 236 с.