Software Product Lines (SPLs) can aid modern ecosystems by rapidly developing large-scale software applications. SPLs produce new software products by combining existing components that are considered as features. Selection of features is challenging due to the large number of competing candidate features to choose from, with different properties, contributing towards different objectives. It is also a critical part of SPLs as they have a direct impact on the properties of the product. There have been a number of attempts to automate the selection of features. However, they offer limited flexibility in terms of specifying objectives and quantifying datasets based on these objectives, so they can be used by various selection algorithms. In this research we introduce a novel feature selection approach that supports multiple multi-level user defined objectives. A novel feature quantification method using twenty operators, capable of treating text-based and numeric values and three selection algorithms called Falcon, Jaguar, and Snail are introduced. Falcon and Jaguar are based on greedy algorithm while Snail is a variation of exhaustive search algorithm. With an increase in 4% execution time, Jaguar performed 6% and 8% better than Falcon in terms of added value and the number of features selected.
A Novel High-Dimensional Trajectories Construction Network based on Multi-Clustering Algorithm
