Probabilistic Algorithms for Election Result Prediction

The paper analyzes the problem of predicting the outcome of elections (how many votes each candidate is going to get), given an imperfect information on the preferences of the voters. We assume that we have a fixed prior on the preferences of each voter for each candidate. We have used two naive algorithms which predict the votes obtained by each candidate in an election. The algorithms are fast and have a linear time and space complexity. We have implemented the algorithms and experimented with simulated data. The results are compared with the method proposed by N. Hazon et al. and it is found to give good results in a very short time. We then experimentally show that these naive algorithms give the same approximation, with linear time complexity.