Hi Ahmed:

how can we make the histories data mining ready?
Should we expand the relation numbers into truth tables
to make it easier for the data mining algorithm to recognize
useful patterns?

Please can you prepare a cd for the histories that is useful for data mining.
It should describe the format that we want to give to Mirek 
to suggest a suitable data mining algorithm for the task.

-- Karl

Using data mining to debug the software

Hope: not all agents make the same mistakes.

What to mine for:

If A2 offers challenge c1 and relation R1 appears in the challenge c1 and c1
is bought by A1, then A1 loses money. Seems to indicate that A1 has a problem
with relation R1.

When A1 buys a challenge using relation R1 from {A2,A3,A4}
and R1's truth table has a one for row 001, then A1 loses money.
Seems to indicate that A2 has a bug related to this truth table entry
while A2 and A3 and A4 do it right.

=============================
Data mining algorithms

1. Model or Pattern Structure
2. Score function: judging the quality of a fitted model
3. Optimizing the score function and searching over different
     model and pattern structures.
4. Data management strategy