Build Machine Learning tool

I want to create a machine learning programme which takes greyhound racing data and predicts the probability of each greyhound winning for future races. I think that the Amazon Machine Learning tool could do this but I am open to other suggestions. To start with I will need to do some data mining to capture the following past data about past races. 1. Trap 2. Race Distance 3. Track 4. Days since last race 5. Split 6. Sex of dog 7. Finishing time 8. Grade of race Suggested sources for this data are: [login to view URL] http://greyhoundbet.racingpost.com/#dog/race_id=1453332&r_date=2016-05-27&dog_id=476762 The output should be represented as a percentage chance of each dog winning, which should always add up to 100%. E.g. Trap 1 10% Trap 2 20% Trap 3 35% Trap 4 5% Trap 5 15% Trap 6 15% I would suggest looking at this link to better understand what I am trying to create. [login to view URL]