Machine learning and artificial intelligence technologies application for software development project efforts (duration) estimation
Jazyk: | ruština |
---|---|
Rok vydání: | 2021 |
Předmět: | |
DOI: | 10.18720/spbpu/3/2021/vr/vr21-4391 |
Popis: | РаÑÑмоÑÑено пÑименение ÑÐµÑ Ð½Ð¾Ð»Ð¾Ð³Ð¸Ð¹ маÑинного обÑÑÐµÐ½Ð¸Ñ Ð¸ иÑкÑÑÑÑвенного инÑеллекÑа к пÑоÑеÑÑÑ Ð¿ÑедваÑиÑелÑного пÑогнозиÑÐ¾Ð²Ð°Ð½Ð¸Ñ Ð´Ð»Ð¸ÑелÑноÑÑи пÑоекÑа по ÑазÑабоÑке пÑогÑаммного обеÑпеÑÐµÐ½Ð¸Ñ Ñ ÑелÑÑ ÑÐ½Ð¸Ð¶ÐµÐ½Ð¸Ñ Ð¾ÑновнÑÑ Ð¿ÑоекÑнÑÑ ÑиÑков. ÐпÑÐµÐ´ÐµÐ»ÐµÐ½Ñ ÑакÑоÑÑ, коÑоÑÑе наиболее ÑилÑно влиÑÑÑ Ð½Ð° длиÑелÑноÑÑÑ ÑÐ°Ð±Ð¾Ñ Ð² подобнÑÑ Ð¿ÑоекÑÐ°Ñ . ÐодобÑÐ°Ð½Ñ Ð°Ð»Ð³Ð¾ÑиÑÐ¼Ñ Ð´Ð»Ñ Ð°Ð½Ñамблевой модели, обеÑпеÑиваÑÑие вÑÑокое каÑеÑÑво пÑогнозов. С помоÑÑÑ ÑзÑка пÑогÑаммиÑÐ¾Ð²Ð°Ð½Ð¸Ñ Python пÑоведено обÑÑение модели и пÑоизведена пеÑекÑеÑÑÐ½Ð°Ñ Ð¿ÑовеÑка ÑезÑлÑÑаÑов пÑогнозиÑÐ¾Ð²Ð°Ð½Ð¸Ñ Ñ ÑелÑÑ Ð¿Ð¾Ð²ÑÑÐµÐ½Ð¸Ñ Ð´Ð¾ÑÑовеÑноÑÑи знаÑений меÑÑик оÑенки каÑеÑÑва. ÐÑоведена комплекÑÐ½Ð°Ñ Ð¾Ñенка каÑеÑÑва полÑÑенной модели Ñ Ð¸ÑполÑзованием полÑÑеннÑÑ Ð´Ð»Ñ Ð½ÐµÐµ знаÑений обÑепÑинÑÑÑÑ Ð¼ÐµÑÑик Ð´Ð»Ñ Ð¾Ñенки моделей маÑинного обÑÑениÑ. ÐÐ¾Ð´ÐµÐ»Ñ Ð±Ñла пÑизнана адекваÑной и обеÑпеÑиваÑÑей вÑÑокÑÑ ÑоÑноÑÑÑ Ð¿Ñогнозов. ÐÑÑÑеÑÑвлено ÑÑавнение полÑÑенной модели Ñ Ð°Ð½Ð°Ð»Ð¾Ð³Ð¸ÑнÑми Ñ Ð¸ÑполÑзованием извеÑÑнÑÑ Ð·Ð½Ð°Ñений меÑÑик каÑеÑÑва. ÐолÑÑÐµÐ½Ñ ÑиÑленнÑе ÑезÑлÑÑаÑÑ, Ñ Ð°ÑакÑеÑизÑÑÑие полÑÑеннÑй вÑигÑÑÑ Ð² ÑоÑноÑÑи пÑогнозов полÑÑенной модели по ÑÑÐ°Ð²Ð½ÐµÐ½Ð¸Ñ Ñ Ð°Ð½Ð°Ð»Ð¾Ð³Ð°Ð¼Ð¸. ÐпÑÐµÐ´ÐµÐ»ÐµÐ½Ñ Ð³ÑаниÑÑ Ð¿ÑименимоÑÑи полÑÑенной модели в оÑганизаÑиÑÑ , занимаÑÑÐ¸Ñ ÑÑ ÑазÑабоÑкой пÑогÑаммного обеÑпеÑениÑ. ÐÐ°Ð½Ñ ÑекомендаÑии по внедÑÐµÐ½Ð¸Ñ Ð¸ иÑполÑÐ·Ð¾Ð²Ð°Ð½Ð¸Ñ Ð¿Ñедложенной модели в Ð¸Ñ Ð¿ÑомÑÑленнÑÑ Ð¿ÑоÑеÑÑÐ°Ñ . The application of machine learning and artificial intelligence technologies for the project duration predicting aimed to avoiding main software development project risks was considered. The factors that have huge influence on the duration of work in such projects have been identified. Algorithms for the ensemble model have been selected that provide high quality forecasts. Using the Python programming language, the model was trained, and the forecasting results were cross-validated in order to increase the reliability of the values of the quality assessment metrics. A comprehensive assessment of the quality of the model was performed using the values of conventional metrics for evaluating machine learning models obtained for it. The model was found to be adequate and fast in forecasting accuracy. A comparison to similar models using the known values of quality metrics was performed. Numerical characteristics were received that characterize the obtained gain in the accuracy of the forecasts of the model in comparison with analogs. The applicability of the model in organizations from software development domain has been determined. Recommendations are given for the implementation and use of the proposed model in their industrial processes. |
Databáze: | OpenAIRE |
Externí odkaz: |