Common Features Based Volunteer and Voluntary Activity Recommendation Algorithm

In general, the dataset of volunteer recommendation systems shows the sparsity, while a volunteer recommendation system required performing the function of recommending voluntary activities interesting to a specific volunteer. To our knowledge, there exists no such kind of recommendation systems. To begin with, this paper firstly presents an analysis of a dataset collected from a real volunteering application website and discovered two features: the locations between the volunteers and the voluntary activities are in close proximity, and the resulting graph which describes the participation relationship between volunteers and voluntary activities is a kind of bipartite, showing many small communities inside it. We call the first discovery ´geographically closely participating´, and the second discovery ´participating together´. Based on these findings, a rating matrix, featuring a matching method for the recommendation algorithm has been constructed. Secondly, we propose a weighted Personal Rank algorithm to implement the required functions of a volunteer recommendation system by employing the registration information of volunteers and voluntary activities. This includes the volunteers´ preferences, activities and location etc. The comparison of proposed method with the rating matrix-based collaborative filter algorithm and the Personal Rank algorithms shows that our proposed method outperforms them.