Option for evaluation of development of decision support system requirements

DOI: 10.31673/2412-9070.2021.014548

Abstract

The quality of the decision support system (DSS) is influenced by the process of creating this system. An important part of the DSS creation process is occupied by events that discuss issues related to the definition of system requirements. Both representatives of the customer and representatives of the executor take part in these actions. The difference between the participants creates uncertainty.
It is important to understand the weaknesses in measures to address system requirements in the early iterations of system development.
This will allow appropriate measures to be taken to improve the quality of measures to address system requirements.
This paper proposes the use of Bayesian methods to evaluate the development of requirements for the creation of a decision support
system.
A model is proposed in which the participants of the events evaluate the measures in terms of addressing all issues and uncertainties.
After the implementation of the requirements discussed at the events, the results of testing are evaluated for these activities. The analysis of the assessments provided by the participants of the activities and evaluations, based on the test results, provides an opportunity to draw appropriate conclusions and take appropriate measures.
During the evaluation process, type I errors occur — the activities were evaluated by the participants on «3», which meant that not all issues were resolved and problems exist, and as a result no errors were made in implementing the precedents worked on at these events. Type II error — the measures were evaluated by the participants on «5», which meant that all issues were resolved and there were no unresolved issues, and as a result errors were obtained in the implementation of precedents, which were worked out at these events.
The article processes the initial data.
The historical representation is determined: P(T|D) = P(D,T)/P(D).
The posterior representation is determined: P(D|T) = P(D,T)/P(T).
Using this mathematical model, we can assess the quality of processing the requirements for the creation of DSS. In the case of obtaining low values of P(D1|T1), P(D2|T2), P(D3|T3) it can be concluded that the measures to process the requirements are not carried out at the appropriate level and may need to be carried out differently. In the case of obtaining low values of P(D3|T1), P(D1|T3) it can be concluded that the measures to process the requirements are carried out at the appropriate level and the probability of errors of I and II kind is quite low. Based on the results of the work on the creation of the first and second stages of DSS, it is possible to draw conclusions and make organizational decisions, and as a result, other stages of the creation of DSS will be better than the first and second stages.
More and more activities are moving to the electronic form, the implementation of the function of estimating the processing of the requirement is becoming easier, so research in this area has prospects.

Keywords: Bayesian methods; decision support systems; evaluation of system requirements development.

References
1. Гаманюк І. М. Варіант застосування байєсівських методів для машинного навчання штучного інтелекту системи підтримання прийняття рішень у боротьбі зі спамом // Зв’язок. 2018. № 6. С. 14–17.
2. Байєсівські мережі в системах підтримання прийняття рішень. Навч. посіб. / М. З. Згуровський, П. І. Бідюк, О. М. Терентьєв, Т. І. Просянкіна-Жарова. Київ: ТОВ «Видавниче Підприємство “Едельвейс”», 2015. 300 с.
3. Горбань І. І. Теорія ймовірностей і математична статистика для наукових працівників та інженерів: монографія. Київ, 2014. 244 с.
4. Chiang S. Jao. Decision Support Systems. Intech 2010. 406 р.