# -*- coding: utf-8 -*-
import dataMngr as dm
import compInfo
def preprocess(fileName, train=True):
df = dm.file2Dataframe(fileName)
inputArr, outputArr = dm.processDataframeForNP(df, train)
inputFileName, outputFileName = dm.makeFileName(fileName)
if train:
dm.save(outputFileName, outputArr)
dm.save(inputFileName, inputArr)
return inputArr, outputArr
print compInfo.dataPoints500File
inA, outA = preprocess(compInfo.dataPoints500File)
print inA.shape
print outA.shape
loadedInA = dm.load(compInfo.outputDataDirectory + 'Innocentive_500_Sample_input.np')
loadedOutA = dm.load(compInfo.outputDataDirectory + 'Innocentive_500_Sample_output.np')
print loadedInA.shape
print loadedOutA.shape
def makeTestOutput(modelfp, inTestFileName, outTestFileName): ML.generateTestOutput(dm.load(modelfp), dm.load(inTestFileName), outTestFileName)
from sklearn.ensemble import RandomForestClassifier
from sklearn.cross_validation import cross_val_score
import dataMngr as dm
# only generated model, doesn't attempt to fit model
def generateModel(model, modelArrParams, modelDictParams):
if modelArrParams is None:
modelArrParams = []
if modelDictParams is None:
modelDictParams = {}
standardModel = model(*modelArrParams, **modelDictParams)
return standardModel
def generateModelScores(stdModel, inData, outData, modelFileName, statisticsFileName):
outData = outData.ravel()
scores = cross_val_score(stdModel, inData, outData)
saveModelAndScores(modelFileName, stdModel, statisticsFileName, scores)
return scores.mean()
def saveModelAndScores(modelFileName, stdModel, statisticsFileName, scores):
dm.save(modelFileName, stdModel)
dm.save(statisticsFileName, scores.mean())
def generateModelAndScores(model, modelArrParams, modelDictParams, inData, outData, numRuns):
for x in xrange(numRuns):
stdModel = generateModel(model, modelArrParams, modelDictParams)
print generateModelScores(stdModel, inData, outData, "RF" + str(x) + ".np", "RF_Stats" + str(x) + ".np")
print generateModelAndScores(RandomForestClassifier, None, None, dm.load("Innocentive_500_Sample_input.np"), dm.load("Innocentive_500_Sample_output.np"), 5)
def generateTestOutput(modelfp, testOutfp, inData): ML.generateTestOutput(dm.load(modelfp), inData, testOutfp)
def checkCrossValidation(inDatafp, outDatafp, modelInfo, numFolds): inData = dm.load(inDatafp) outData = dm.load(outDatafp).ravel() model = generateUnfittedModel(modelInfo["model"], modelInfo["modelArrParameters"], modelInfo["modelDictParameters"]) return cv.cross_val_score(model, inData, outData, cv=numFolds)
