I want to predict the voting behavior of american politicians. I believe that when people read different bills, they make a decision in different topics of a bill, just like following a decision tree. But I don't know how to combine the decision tree with their behavior? Can anyone give me some advice? Thank you!
It looks like to build a classification model. Many method can be used for classification, like support vector machine, artificial neural network etc. I think the input of the model can be properties of people and bills, such as sex, age (can be define in some range, like 1-10, 1-15, 15-20, 20-30 etc), country, type of the bill etc.. The output of the model is voting behaviour that you redefined.
It is a classification application as mentioned in the other answers. There is an uncertainty nature in the voting therefore, decision trees are not practical. For example, in decision trees if a candidate is over 60 years old, he has to be either 100% out all the time, or 100% selected. This level of certainity for the important criteria does not exist in the real world.
Instead, there are many classification methods that can classify a set of data with some degree of certainty. Suggesting one classifier and promising that it will work for any classification applications is not feasible. Linear classifiers, support vector machines, artificial neural networks, etc. are among the methods that you might want to go through them and see which one gives you more precise predictions on the previous results. It's like asking us which vehicle is the best for transportation. Cars are good for traveling but for more than 30 people you need a bus. If you want to start, I would recommend artificial neural networks. There are many types of them. They can be trained based on the training data that you provide. They will adjust themselves (the weights, biases, etc.) to serve as an adaptive predictor.
Good luck
I agree with my colleagues,
A good approach is using a classification model. My suggestion is modeling individual vote participant and modeling his vote like function of his historic and current vote of other participants, of course maybe you must add other features like issues that they are voting.
Regards,
Danilo Reis
- Badges
- Science topic