Algorithms and Architectures for Parallel Processing by Guojun Wang Albert Zomaya Gregorio Martinez Perez & Kenli Li

Author:Guojun Wang, Albert Zomaya, Gregorio Martinez Perez & Kenli Li
Language: eng
Format: epub
Publisher: Springer International Publishing, Cham

4.2 Clustering Method

Our clustering method periodically groups workers into k clusters and each cluster consists of workers with similar characteristics.

In the above algorithm, each worker and center is both represented by a vector. Vector of worker consists of all skill degree while center is the average value in each dimension, calculated by workers belongs to this cluster. distance() is the Euclidean distance between two vectors. MAXIMUM is used to control the convergence of algorithm to avoid high time complexity caused by discrete data. For the first time clustering, initial parameters are selected randomly. Following processes can take previous results as inputs to improve efficiency.

Besides, since the number of skills related to software development is very large, it is redundant to use a single vector to record all skill degree for each worker. Thus, as in the case of Upwork, we classify tasks into different categories and remove unnecessary skills in this category to avoid the sparsity and cold-start problem in recommending system [10]. Considering different cognitions against classification, we don’t specify categories in this paper and each crowdsourcing platform can build their own rules of classification.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.