# Algorithm to predict next 3D points

For example, having this data:

year x/y/z
2007 10/20/70
2008 20/10/70
2009 30/10/60
2010 40/10/50
2011 40/15/45


We want to predict what will be the x/y/z in 2012. We can observe that x is constantly increasing, y rather constant, z decreasing, so we can predict, for example:

2012 50/10/40


How to predict that x/y/z values?

x/y/z refers to number of positive/neutral/negative sentiments in given year. We always get 100 sentiments of news for a year. In our example, in year 2007 there were 10 positive, 20 neutral and 70 negative sentiments. So, x/y/z are in range <0; 100> and x + y + z = 100. x/y/z seems to be independent. Number of consecutive years from which we are gathering data is in range <2; 10>. In example it was 5. Algorithm is going to be implemented in Python.

-
Have you looked at linear regression? en.wikipedia.org/wiki/Linear_regression – Adam Dec 27 '11 at 20:10
For a split second, I thought you were the same Adam. – Raskolnikov Dec 27 '11 at 21:39