Low-rank approximations for predicting voting behaviour
نویسندگان
چکیده
Online voting networks, the ones where users can give a positive or negative evaluation (as friend or foe respectively) to their peers, have caught the eye of the Data Mining community. Researchers have come up with some ideas on how to predict their behavior based on ad hoc features, mainly calculated by aggregation of their local vecinity data. We introduce a novel approach for predicting social networks behavior by modeling the tasks as a model based recommender system problem. Both formulations are very simmilar. Recommender Systems deals with records of users ratings for products, while voting networks consits of evaluations given by users for one another. The former has a broad body of literature that deals with many problems found in social networks, such as: the curse of dimensionality, sparsity and cold start. Moreover, its development has been heavily pushed producing very robust and sound models. We show that our method achieves similar or better performance than current state of the art methods for the Link Prediction and Polarity Prediction tasks. Finally we further extend our approach to include nodes and edges attributes, other than the basic graph structure, as sample features.
منابع مشابه
A Study of Factors that Influenced Voters’ Behaviour: A Case of Billiri Local Government (2019 General Elections)
متن کامل
A New Voting Model For Groups With Members of Unequal Power and Proficiency
To proposing a voting model for groups with members of unequal power and proficiency, we present some models for rank ordering efficient candidates, by extending the ideas of some authors. Then, we propose a new methodology to rank the ranking models for the performance indices of only DEA efficient candidates based on a classical voting model. Also, an approach for combining the results obtain...
متن کاملVoting and different ranking orders in DEA
Data Envelopment Analysis (DEA) technique now widely use for efficiency evaluation of a set of Decision Making Units (DMUs). As regards of the necessity for ranking efficient units different DEA models presented each of which has advantages and rank efficient units from special aspects. Note that all the existing ranking models have disadvantages, as well and there is not a model in which all t...
متن کاملLow-Rank Approximations with Sparse Factors I: Basic Algorithms and Error Analysis
We consider the problem of computing low-rank approximations of matrices. The novel aspects of our approach are that we require the low-rank approximations be written in a factorized form with sparse factors and the degree of sparsity of the factors can be traded oo for reduced reconstruction error by certain user determined parameters. We give a detailed error analysis of our proposed algorith...
متن کاملA two phases approach for discriminating efficient candidate by using DEA inspired procedure
There are several methods to ranking DMUs in Data Envelopment Analysis (DEA) and candidates in voting system. This paper proposes a new two phases method based on DEA’s concepts. The first phase presents an aspiration rank for each candidate and second phase propose final ranking.
متن کامل