Predicting future successful teams of experts that can work efficiently and effectively together is a challenge due to the nature of the experts who may evolve over time and that their skill sets aren't static. Previous methods often overlook capturing the temporal aspects of experts' skills over time and there's a need for improvement in this regard that accounts for time.
Proposed method
In this paper, the author proposes two main components: streaming training strategy and temporal information as an additional signal. The first component acts as a way to inform the learning process of past collaborations of the experts. The second component adds another input stream, time, to the pipeline which helps with predicting future teams. This idea is to produce more accuracy in prediction with the added temporal component.
My Summary
The proposed strategy showed improvements in the results over the four datasets that were used for testing. The proposed method improves prediction power in neural models, thus neural models that utilize the author's temporal information strategy outperformed the neural baseline models. However, there was some room for improvements that could be addressed in future works.
Link: https://link.springer.com/chapter/10.1007/978-3-031-56027-9_20
Main problem
Predicting future successful teams of experts that can work efficiently and effectively together is a challenge due to the nature of the experts who may evolve over time and that their skill sets aren't static. Previous methods often overlook capturing the temporal aspects of experts' skills over time and there's a need for improvement in this regard that accounts for time.
Proposed method
In this paper, the author proposes two main components: streaming training strategy and temporal information as an additional signal. The first component acts as a way to inform the learning process of past collaborations of the experts. The second component adds another input stream, time, to the pipeline which helps with predicting future teams. This idea is to produce more accuracy in prediction with the added temporal component.
My Summary
The proposed strategy showed improvements in the results over the four datasets that were used for testing. The proposed method improves prediction power in neural models, thus neural models that utilize the author's temporal information strategy outperformed the neural baseline models. However, there was some room for improvements that could be addressed in future works.
Datasets