National Election Prediction Lei Xu
Motivation 2016 president election is coming. Really useful for politicians who want to know which factors may determine a voter's choice. And I was fascinated by a TV series....
Problem to Solve Given 4 characteristics of an individual voter,using artificial neural network to classify their voting choice. Data set contains 2000 individual-level turnout data. It pools several American National Election Surveys conducted during the 1992 presidential election year. 2,000 observations are included in the sample data.
Data Set continue In the original data ,each table containing 5 variables ("race", "age", "educate", "income", and "vote"). 4 features: race educate income vote sample data race age educate income vote white 60 14 3.3458 1
Approach Config the ANN sturture in MATLAB by modifiying Hesham Eraqi's code Sigmoid activation function Backpropagation training use mean square error(MSE) to messure the performance stops when MSE =0 or exceed the maximum epoch value
Results MSE=0.020619 [10 10]; η:0.15; [10 10]; η:0.01; MSE=0 [10 10]; η:0.15; [10 10]; η:0.01; MSE=0 [10 10]; η:0.10 MSE=0
[10 10]; η:0.15; MSE=0.020619 [10 10 10] η=0.15 MSE=0.1856 [5 5] η=0.15 MSE=0.39691
Discussion The simpler the better job, but not always. The learning rate η doesn't impact the performance too much using the ANN to analyze the whole country voter data ,we may predict the final result of the election. limitation Future development
Reference King, Gary, Michael Tomz, Jason Wittenberg (2000). “Making the Most of Statistical Analyses: Improving Interpretation and Presentation,” American Journal of Political Science, vol. 44, pp.341–355. http://heraqi.blogspot.com.eg/2015/11/mlp-neural-network-with-backpropagation.html http://neuralnetworksanddeeplearning.com/chap3.html Lecture Slides
Thanks