[Tutor] Program to Predict Chemical Properties and Reactions

[Dave Angel]

On 07/16/2011 05:32 PM, B G wrote:
> Thanks, Emile-- although I'm not sure I was completely clear about my
> objective. What I really meant is that is there a way (via machine learning)
> to give the computer a list of rules and exceptions, and then have it
> predict based on these rules the ionization energy. Ultimately I'm pretty
> interested in the idea of building a program that could predict the expected
> result of a chemical reaction, but that's a huge project, so I wanted to
> start with something much, much, much smaller and easier.
>
> So I don't really want to manually input ionization energy, atomic radius
> size, etc, since I'm not sure how much that would help towards the bigger
> goal.  For this project, I already have the ~15 rules that describe trends
> on the periodic table.  I want to write a program that, based on these rules
> (and the few exceptions to them), could predict the inputs that I want.
>   Does this make sense?
>
>
Neither ordering nor trends will give values for individual items.  So 
unless these rules are much more numerical than the clues you've 
mentioned so far, the problem is clearly impossible.

But I suspect the problem is possible, and that you have much more 
information than you intend to tell us.

-- 

DaveA