Application of machine learning techniques

[dolejarz]

Tentative game plan:

• Create a table of 90% of rows that have a MIR field
• Train a decision tree classifier using nLanes, incType, and incDuration as features, MIR as labels
• Test accuracy against the remaining 10% of rows
• If accuracy is good, go ahead and label the rest of the rows that don't have a MIR field

[dolejarz]

using pd.df in sklearn remember to cast strings to int/float!!!

[dolejarz]

use pickle to pass numpy arrays between preprocessing script and machine learning script

[dolejarz]

Tweak parameters of decision tree classifier to get higher precision and recall

[dolejarz]

use sklearn's one hot encoder in preprocessing. Use label encoder for non numerical values