Open source software (OSS) has a large number of users, and as their number increases, the effects of OSS failures also increase. This thesis research conducted clustering based on the shape of the graph of development processes of OSS projects. After clustering, the OSS success factors suggested in previous studies were analyzed cluster-by-cluster to identify clusters that had many of these successful projects. Then, the success factors proposed were analyzed cluster-by-cluster to determine which had real effects.
This study used the top 5,000 GitHub projects and divided them into four clusters. Of the four clusters, the most successful, sustainable projects were those in which development began in the early stage and progressed slowly over time. The projects in this cluster were more likely to operate on organizational accounts that used project names, and they attracted developers’ attention by using dedicated homepages and search keyword tags. They also had a large number of core developers, few pending issues, and more releases than the other clusters.