.png){kind=icon}
.png){kind=icon}
Return
Developer Recommendation and Team Formation in Collaborative Crowdsourcing Platforms
.png){kind=icon}
OA
AI
DOI：10.1109/ACCESS.2025.3558557.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The Competitive Crowdsourcing Software Development (CSE) environment is a dynamic and evolving field that interests researchers and the software industry. This ecosystem ensures the timely delivery of cost-effective, innovative, and high-quality solutions. As this environment grows in popularity, it faces the crucial challenge of developer recommendation and team formation, which must be addressed to ensure its ongoing success and progress. This research offers a novel method for assembling teams and recommending developers. It combines keyword-based embeddings and fuzzy logic to match developers with appropriate tasks with exceptional accuracy. The technique utilizes KeyBERT to extract keywords and perform embedding to capture relevant skills and task requirements without emphasizing common words. The embeddings are optimized using a fuzzy logic framework. This framework categorizes the quality of developer-task pairings into three distinct levels: Strong, Average, and Weak. This approach enables developer recommendations and team formation, balancing precision, recall and F1-score across varying team sizes to achieve high overall accuracy by a percentage increase of 4.19%. and TF score increased by 4.05%. The proposed method consistently outperforms existing methods, allowing the formation of capable and reliable teams of varying sizes. This ensures the creation of high-performing, well-balanced teams that can effectively handle diverse tasks with the percentage improvement from Content-Based Filtering (CBF). The results in percentage increase (precision, recall and f1-score) of the proposed approach (with k=5) from CBF improve by 30.29%, 36.49%, and 33.45%, and with User-based Collaborative Filtering (UCF) and Random Recommendation (RR) performances increase by (19.58%, 15.98%, and 17.92%) and (160.17%, 215.07%, and 188.31%), respectively.
Keywords:
Collaboration
Fuzzy logic
Software
Industries
Crowdsourcing
Semantics
Electronic mail
Faces
Complexity theory
Accuracy
CSE
Topcoder
key-bert
developer recommendation
NLP
team formation
fuzzy logic
classification
Journal
IF:
3.6
Papers:
9.8W
Citations:
29.4W