How to classify the opinions to get the best prediction?

Question

Assume I have four friends, each of them will predict the result of each football match (based on history data, weather data, etc). They will tell me their decisions.

I also have the set of their data history data and weather data. I can also do the similar prediction. However, I want to classify which of my four friend make the right decision instead of predict the football match. (The reason I said that because I do NOT want to put all the data including my friends' decisions into a classify to make the decision. It is because their decisions will be the same weight of the history and weather data. My friends' decisions should weight much more )

How I can maximize the correct prediction?

Answer 1

Your model is highly inefficient. Each of the players prediction has no influence on the other players predictions' truthfulness. So instead of inputting them all at once, you should input them sequentially. So:

friend1 + other data = output
friend2 + other data = output
friend3 + other data = output
friend4 + other data = output

This makes your network smaller, making the training happen in less iterations and in a less amount of time.

But still. You shouldn't input your friends prediction at all. The only thing that your network is doing is comparing it's own prediction with your friends prediction internally.

In what way should your friends' predictions add anything useful to the (each others) output?