foldId] = evaluation.getAverageOperationCosts(
testLabels, predictedTestLabels, avgTestFeatureCosts,
falsePositiveCost)
print("NEW ULTRA FAST VERSION")
print("dataName = ", dataName)
print("NUMBER_OF_FOLDS = ", constants.NUMBER_OF_FOLDS)
print("onlyLookingOneStepAhead = ", onlyLookingOneStepAhead)
print("densityRegressionModelName = ", densityRegressionModelName)
print("falseNegativeCost = ", falseNegativeCost)
print("RESULTS WITH DYNAMIC ACQUISTION (SLOW VERSION): ")
print(
"*************************** AVERAGE OVER ALL FOLDS *******************************************"
)
evaluation.showHelper("total costs = ", testTotalCostsAllFolds)
evaluation.showHelper("feature costs = ", testFeatureCostsAllFolds)
evaluation.showHelper("misclassification costs = ",
testMisClassificationCostsAllFolds)
evaluation.showHelper("accuracy = ", testAccuracyAllFolds)
evaluation.showHelper("AUC = ", testAUCAllFolds)
resultsRecorder.addAll(
falsePositiveCost,
(testTotalCostsAllFolds, testFeatureCostsAllFolds,
testMisClassificationCostsAllFolds, testAccuracyAllFolds,
testAUCAllFolds, testRecallAllFolds, testFDRAllFolds,
testOperationCostsAllFolds, testRecallAllFolds_exactRecall,
testFDRAllFolds_exactRecall,
testOperationCostsAllFolds_exactRecall))
print("variationName = ", variationName)
print("*************************** GREEDY MISER *******************************************")
totalCostsValidResult, validAccuracy, validFeatureCosts, bestLambdaId, bestTreeId, avgBestTotalCostsValidResult = evaluation.getBestAverage10TrainFoldTotalCostsResultSymmetricForGreedyMiser(dataName, testFoldId, misclassificationCostsSymmetric, sameClassCost)
allTestFoldAvgBestTotalCostsValidResult[testFoldId] = avgBestTotalCostsValidResult
allBestSettings[testFoldId, 0] = bestLambdaId
allBestSettings[testFoldId, 1] = bestTreeId
validTotalCostsAllFolds[testFoldId] = totalCostsValidResult
validFeatureCostsAllFolds[testFoldId] = validFeatureCosts
validMisClassificationCostsAllFolds[testFoldId] = (1.0 - validAccuracy) * misclassificationCostsSymmetric + validAccuracy * sameClassCost
validAccuracyAllFolds[testFoldId] = validAccuracy
# print("bestLambdaId = ", bestLambdaId)
print("*************************** AVERAGE OVER ALL FOLDS (misclassification costs = " + str(misclassificationCostsSymmetric) + ", sameClassCost = " + str(sameClassCost) + ") *******************************************")
evaluation.showHelper("total costs = ", validTotalCostsAllFolds)
evaluation.showHelper("feature costs = ", validFeatureCostsAllFolds)
evaluation.showHelper("misclassification costs = ", validMisClassificationCostsAllFolds)
evaluation.showHelper("accuracy = ", validAccuracyAllFolds)
evaluation.showHelper("average best validation costs (for analysis only) = ", allTestFoldAvgBestTotalCostsValidResult)
# save best settings from validation data
outputFilename = "/export/home/s-andrade/dynamicCovariateBaselines/GreedyMiser/" + dataName + "_" + str(int(misclassificationCostsSymmetric)) + "_allBestSettings"
matlabDict = {}
matlabDict["allBestSettings"] = numpy.asmatrix(allBestSettings, dtype = numpy.int)
scipy.io.savemat(outputFilename, matlabDict)
if os.path.isfile(experimentSetting.MATLAB_DATA_FOLDER_RESULTS + "greedyMiser/" + dataName + "_" + str(int(misclassificationCostsSymmetric)) + "_forFinalTrainingAndTesting_" + str(4) + "_allResults.mat"):