def MLPRegressorPredict(playerFirstName, playerLastName, target, batter = False, solver='lbfgs'):
statMap = pitchStatMap
if batter:
statMap = batStatMap
if target not in statMap.keys():
print "bad target value"
data2016, null, playerData = dataUtil.getPlayerInformation(playerFirstName + " " + playerLastName)
features = statMap[target]
x = []
y = []
calculate2015data = []
# print playerFirstName, playerLastName, playerData
if not playerData or '2015' not in playerData:
return None
for i, year in enumerate(playerData):
if year == '2015':
continue
curArray = []
if not playerData.get(str(int(year) + 1), 0):
continue
for feature in features:
curArray.append(float(playerData[year][feature]))
x.append(curArray)
y.append(float(playerData[str(int(year) + 1)][target]))
for feature in features:
calculate2015data.append(float(playerData['2015'][feature]))
if not x or not y:
return None
regr = MLPRegressor(hidden_layer_sizes=(1000,),solver=solver)
regr.fit(x, y)
return regr.predict(calculate2015data)[0]
def predict(playerFirstName, playerLastName, target, batter = False):
statMap = pitchStatMap
if batter:
statMap = batStatMap
if target not in statMap.keys():
print "bad target value"
data2016, null, playerData = dataUtil.getPlayerInformation(playerFirstName + " " + playerLastName)
features = statMap[target]
x = []
y = []
calculate2016data = []
for year in playerData:
curArray = []
for feature in features:
curArray.append(float(playerData[year][feature]))
x.append(curArray)
y.append(float(playerData[year][target]))
for feature in features:
calculate2016data.append(float(data2016[feature]))
regr = linear_model.LinearRegression()
regr.fit(x, y)
return regr.predict(calculate2016data)[0]
def SVRPredict(playerFirstName, playerLastName, target, batter = False, kernel='rbf', C=1.0, epsilon=0.1, myFeatures=None):
statMap = pitchStatMap
if batter:
statMap = batStatMap
if target not in statMap.keys():
print "bad target value"
data2016, null, playerData = dataUtil.getPlayerInformation(playerFirstName + " " + playerLastName)
if myFeatures == None:
features = statMap[target]
else:
features = myFeatures
x = []
y = []
calculate2015data = []
# print playerFirstName, playerLastName, playerData
if not playerData or '2015' not in playerData:
return None
for i, year in enumerate(playerData):
if year == '2015':
continue
curArray = []
if not playerData.get(str(int(year) + 1), 0):
continue
for feature in features:
curArray.append(float(playerData[year][feature]))
x.append(curArray)
y.append(float(playerData[str(int(year) + 1)][target]))
for feature in features:
calculate2015data.append(float(playerData['2015'][feature]))
if not x or not y:
return None
regr = SVR(kernel=kernel, C=C, epsilon=epsilon)
regr.fit(x, y)
return regr.predict(calculate2015data)[0]
def nextYearPredict(playerFirstName, playerLastName, target, batter=False):
statMap = pitchStatMap
if batter:
statMap = batStatMap
if target not in statMap.keys():
print "bad target value"
data2016, null, playerData = dataUtil.getPlayerInformation(
playerFirstName + " " + playerLastName)
features = statMap[target]
x = []
y = []
calculate2015data = []
if not playerData or '2015' not in playerData:
return None
for i, year in enumerate(playerData):
if year == '2015':
continue
curArray = []
if not playerData.get(str(int(year) + 1), 0):
continue
for feature in features:
if feature == "G":
indexOfG = features.index(feature)
if float(playerData[year][feature]) == 0:
curArray.append(float(playerData[year][feature]) + 1)
else:
curArray.append(float(playerData[year][feature]))
curArray = normalize(curArray, indexOfG)
if curArray == None:
print playerFirstName, playerLastName, target, features
x.append(curArray)
y.append(float(playerData[str(int(year) + 1)][target]))
index = -1
for feature in features:
if feature == 'G':
games = float(playerData['2015'][feature])
index = features.index(feature)
calculate2015data.append(float(playerData['2015'][feature]))
if not x or not y:
return None
regr = linear_model.LinearRegression()
regr.fit(x, y)
prediction = regr.predict(normalize(calculate2015data, index))
val = prediction[0]
if val <= 0:
return 0.1
return val
def nextYearPredict(playerFirstName, playerLastName, target, batter = False):
statMap = pitchStatMap
if batter:
statMap = batStatMap
if target not in statMap.keys():
print "bad target value"
data2016, null, playerData = dataUtil.getPlayerInformation(playerFirstName + " " + playerLastName)
features = statMap[target]
x = []
y = []
calculate2015data = []
# print playerFirstName, playerLastName, playerData
if not playerData or '2015' not in playerData:
return None
for i, year in enumerate(playerData):
if year == '2015':
continue
curArray = []
if not playerData.get(str(int(year) + 1), 0):
continue
for feature in features:
curArray.append(float(playerData[year][feature]))
x.append(curArray)
y.append(float(playerData[str(int(year) + 1)][target]))
for feature in features:
calculate2015data.append(float(playerData['2015'][feature]))
if not x or not y:
return None
regr = linear_model.LinearRegression()
regr.fit(x, y)
prediction = regr.predict(calculate2015data)
if playerFirstName == 'Nolan' or prediction[0] < 0:
print playerFirstName + " " + playerLastName, target
print regr.coef_
print features
print x
print y
print calculate2015data
print prediction
val = prediction[0]
if val <= 0:
return 0.1
return val
