def saveModel(model, filename):
# Change to where you want to save the model
setCurrentWorkingDirectory('SavedModels')
with open(filename, 'wb') as file:
pickle.dump(model, file)
def makeInterpretPredictions(currentDate, season, startOfSeason):
setCurrentWorkingDirectory('SavedModels')
print('Predictions for ' + currentDate + ':')
predictions = predictDailyGames(currentDate, season, startOfSeason)
interpretPredictions(predictions)
def createModel(startYear=None, startMonth=None, startDay=None, endYear=None, endMonth=None, endDay=None):
# allGames = getTrainingSet(startYear, startMonth, startDay, endYear, endMonth, endDay) # Unnecessary if using data from CSV file
# allGamesDataframe = createDataFrame(allGames) # Unnecessary if using data from CSV file
setCurrentWorkingDirectory('Data')
allGamesDataframe = pd.read_csv('games.csv') # Should be commented out if needing to obtain data on different range of games
logRegModel = performLogReg(allGamesDataframe)
saveModel(logRegModel)
def getTrainingSetCSV(startYear,
startMonth,
startDay,
endYear,
endMonth,
endDay,
season,
startDateOfSeason,
filename='gamesWithInfo.csv'):
# Gets date, teams, and z-score difs for every game within range
rangeOfGames = getTrainingSet(startYear, startMonth, startDay, endYear,
endMonth, endDay, season, startDateOfSeason)
rangeOfGamesDataframe = createDataFrame(rangeOfGames)
setCurrentWorkingDirectory('Data')
rangeOfGamesDataframe.to_csv(filename)
def createModel(startYear=None,
startMonth=None,
startDay=None,
endYear=None,
endMonth=None,
endDay=None,
season='2018-19',
startOfSeason='10/16/2018',
filename='rfModel.pkl'):
# allGames = getTrainingSet(startYear, startMonth, startDay, endYear, endMonth, endDay, season, startOfSeason) # Unnecessary if using data from CSV file
# allGamesDataframe = createDataFrame(allGames) # Unnecessary if using data from CSV file
setCurrentWorkingDirectory('Data')
allGamesDataframe = pd.read_csv(
'COMBINEDgamesWithInfo2016-19.csv'
) # Should be commented out if needing to obtain data on different range of games
rfModel = performRF(allGamesDataframe)
saveModel(rfModel, filename)
def getPredictionsCSV(gameDataFilename, outputFilename):
setCurrentWorkingDirectory('Data')
gamesWithZScoreDifs = pd.read_csv(gameDataFilename)
withoutNums = gamesWithZScoreDifs.loc[:, 'Home':
'Date'] # Slices dataframe to only includes home through date
print(withoutNums)
justZScoreDifs = gamesWithZScoreDifs.loc[:, 'W_PCT':
'TS_PCT'] # Slices dataframe to only include statistical differences
setCurrentWorkingDirectory('SavedModels')
with open('model12.pkl', 'rb'
) as file: # Change filename here if model is named differently
pickleModel = pickle.load(file)
predictions = pickleModel.predict(
justZScoreDifs) # Creates list of predicted winners and losers
probPredictions = pickleModel.predict_proba(
justZScoreDifs) # Creates list of probabilities that home team wins
numCorrect = 0
numWrong = 0
allGames = []
for i in range(len(probPredictions)):
winProbability = probPredictions[i][1]
homeTeam = withoutNums.iloc[i, 0]
awayTeam = withoutNums.iloc[i, 1]
date = withoutNums.iloc[i, 10]
currentGameWithPred = [date, homeTeam, awayTeam, winProbability]
allGames.append(currentGameWithPred)
# Creates dataframe that holds all games info and predictions
predictionsDF = pd.DataFrame(
allGames,
columns=['Date', 'Home', 'Away', 'Home Team Win Probability'])
setCurrentWorkingDirectory('Data')
predictionsDF.to_csv(
outputFilename
) # Saves game info with predictions in data folder as csv file
value = withoutNums.iloc[i, 9]
if value == predictions[i]:
numCorrect += 1
else:
numWrong += 1
print('Accuracy:')
print((numCorrect) / (numCorrect + numWrong)
) # Prints accuracy of model in predicting games for specified range
def makeInterpretPredictions(currentDate, season, startOfSeason):
setCurrentWorkingDirectory('SavedModels')
predictions = predictDailyGames(currentDate, season, startOfSeason)
return (interpretPredictions(predictions))