Models complexity measurement

Abstract: The demand for measuring the quality aspects and need for higher maintainability and understandability of the models are increasing within the field of software engineering and management. Among these, complex models are of special interest for designers as they are more correlated to the eventual reliability of the system and therefore are considered very important. This study presents a method for measuring the complexity of existing software models in Ericsson seeking to raise the maintainability and understandability of the software engineering project in progress. A literature survey was performed in order to find a list of all potentially useful metrics. Narrowing down the long list of metrics was carried out by interviews with designers at Ericsson. Utilizing statistical data analysis based on interviews results was the next step. Beside, workshops were used for evaluating the reliability of preliminary data analysis and an empirical formula was generated for models’ complexity prediction. Metrics such as “non-self-transitions”, “transitions per states”, and “state depth” are the most important for calculating the models’ complexity score (rank) and for these metrics threshold values were set. Challenges and experiences gained in this study demonstrated the importance of incorporating user generated feedback in the empirical complexity modeling studies