def experiment(train_file, test_file, predictor, ntrees, mtry, sample, ignores,
parser_header, cpu_cores):
# Load train data
trainDataX = pandas.read_csv(train_file, header=0)
fmap = preprocessFrame(trainDataX)
predictorCol = trainDataX.columns[predictor]
trainDataY = array(trainDataX.pop(predictorCol))
trainDataX = array(popIgnores(trainDataX, ignores))
timeTrainRF = time()
rf = WiseRF(n_estimators=ntrees, n_jobs=cpu_cores)
rf.fit(trainDataX, trainDataY)
timeTrainRF = time() - timeTrainRF
# Validation
testDataX = pandas.read_csv(test_file, header=0)
preprocessFrame(testDataX, fmap)
predictorCol = testDataX.columns[predictor]
testDataY = array(testDataX.pop(predictorCol))
testDataX = array(popIgnores(testDataX, ignores))
timeTestRF = time()
predict = rf.predict(testDataX)
print predict
testScore = rf.score(testDataX, testDataY)
timeTestRF = time() - timeTestRF
print """
Train time: {0}
Test time: {1}
Err. rate: {2} %
Trees,Sample,Mtry,TrainTime,TestTime,OOB,ClassErr
{3},NA,NA,{0},{1},NA,{4}""".format(timeTrainRF, timeTestRF,
100 * (1 - testScore), ntrees,
1 - testScore)
