РАННЄ ОЦІНЮВАННЯ КІЛЬКОСТІ РЯДКІВ КОДУ ВЕБ-ЗАСТОСУНКІВ, ЩО СТВОРЮЮТЬСЯ ЗА ДОПОМОГОЮ PHP ФРЕЙМВОРКІВ

The problem of early estimation of lines of code count in software projects holds significant importance, as it directly influences the prediction of software development effort, including web applications created using the well-known PHP frameworks, as the CakePHP and Codeigniter. The object of the study is the process of early estimating the lines of code count of web applications created using the CakePHP and Codeigniter frameworks. The subject of the study is the regression models for early estimating the lines of code count of web applications created using the CakePHP and Codeigniter frameworks. The goal of the work is to build some regression models with three factors for early estimating the lines of code count of web applications created using the CakePHP and Codeigniter frameworks depending on the factors that can be found in the class diagram. In the work, we built one nonlinear and two linear regression models for early estimating the lines of code count of web applications created using the CakePHP and Codeigniter frameworks depending on three factors: the number of classes, the average number of methods per class, metric DIT (Depth of Inheritance Tree) on the application level. These factors were chosen for two reasons. First, the values of these factors can be found from the class diagram, and, second, there is no problem of multicollinearity for them. The parameter estimates of the obtained models were found by the method of least squares. The above models are constructed based on splitting the four-dimensional data set (the actual size in thousands of lines of code, the number of classes, the average number of methods per class, the DIT metric on the application level) into two clusters. This data of metrics for 88 open-source web applications created using the CakePHP and Codeigniter frameworks was obtained using the PhpMetrics tool. These applications are hosted on the GitHub platform. The comparison of the constructed models with the other non-linear regression models is performed. The constructed regression models, in comparison with existing non-linear ones, allow us to describe all the four-dimensional data on which they were built.