def run():
# from get_data import get_data
# data,compare_data,_=get_data()
from params import j, term
for i in range(term):
from get_data import get_data
data, compare_data, _ = get_data()
j += rmse(GAIN(data), compare_data)
print(j / term)
def run():
data,compare_data=get_data()
from params import j,term
for i in range(term):
data,compare_data=get_data()
X_train, x_test, Y_train, y_test = train_test_split(data,compare_data,random_state=0)
lr=lr_emsemble_train(X_train,Y_train)
j+=rmse(lr_emsemble_predict(x_test),y_test)
print(j/term)
def run():
from params import j, term
for i in range(term):
from get_data import get_data
data, compare_data, _ = get_data()
d = em(data)
from pandas import DataFrame
d = DataFrame(d,
index=compare_data.index,
columns=compare_data.columns)
for index, col in enumerate(compare_data):
d[col] = d[col].astype(str(compare_data[col].dtype))
j += rmse(d, compare_data)
print(j / term)
def run():
mark = []
from params import j, term
for i in range(term):
data, compare_data, _ = get_data()
d = data.copy()
mask = data.isnull()
model = train(data.fillna(0).values, mask, 30)
filled_data = mida(model, data.fillna(0).values)
from pandas import DataFrame
filled_data = DataFrame(filled_data,
index=compare_data.index,
columns=compare_data.columns)
for index, col in enumerate(compare_data):
filled_data[col] = filled_data[col].astype(
str(compare_data[col].dtype))
j += rmse(filled_data, compare_data)
mark.append(rmse(filled_data, compare_data))
# for i,j in enumerate(mark):
# print(i+2," err:",j)
print(mark)
print(np.sum(mark) / len(mark))
def best_test(op="RMS"):
best_len1 = 5
best_len2 = 10
best_loss = 999999999999.
for len2 in range(5, 10):
for len1 in range(len2 + 1, len(data.columns)):
d = test(len1, len2, data, op)
# print(d)
for col in compare_data.columns:
d[col] = d[col].astype(str(compare_data[col].dtype))
#print(d)
# input()
t_loss = rmse(d, compare_data)
if (float(t_loss) < best_loss):
best_len1 = len1
best_len2 = len2
best_loss = t_loss
return best_len1, best_len2, best_loss
def netflix_eval (probeFile, movieIDYear, custDecadeAvgRatings, movieIDAvgRating, custIDAvgRating, actualRatings) :
"""
Applies heuristics to predict ratings and calculates the RMSE
probeFile is the path to probe.txt from the command line
movieIDAvgRating is the dictionary of {movie ID, average rating)
custIDAvgRating is the dictionary of (cust ID, average rating)
actualRatings is the list of actual customer ratings for RMSE calculation
"""
movieIDpredRatings = {} # {movieID:[ratings]} for printing
allPredRatings = []
with open(probeFile, 'r') as f_myfile:
lines = f_myfile.readlines()
movieID = ""
for line in lines :
if re.search(':', line) : #movieID
predRatings = []
movieID = line.strip(':\r\n')
else :
assert movieID
custID = line.strip() #strip newline
#look up year
year = movieIDYear[movieID]
#determine the decade
decade = netflix_decade_calc(year)
if decade in custDecadeAvgRatings[custID]:
custDecadeRating = float(custDecadeAvgRatings[custID][decade])
else : #use individual customer average if customer didn't rate any movies in that decade
custDecadeRating = float(custIDAvgRating[custID])
pred = (.4*float(movieIDAvgRating[movieID]) + .6*custDecadeRating) #RMSE = 0.971
assert type(pred) is float
predRatings.append(pred)
allPredRatings.append(pred)
movieIDpredRatings[movieID] = predRatings
answer_rmse = rmse(actualRatings, allPredRatings)
return [answer_rmse, movieIDpredRatings]
def run():
data, compare_data = get_data()
data = compare_data
rows, cols = data.shape
shuffled_index = np.random.permutation(rows)
train_index = shuffled_index[:int(rows * (1 - test_size))]
test_index = shuffled_index[int(rows * (1 - test_size)):]
train_data = data.values[train_index, :]
test_data = data.values[test_index, :]
scaler = MinMaxScaler()
scaler.fit(train_data)
train_data = scaler.transform(train_data)
test_data = scaler.transform(test_data)
data, mask = missing_method(test_data, mechanism, method, miss_rato)
missed_data = torch.from_numpy(data).double()
train_data = torch.from_numpy(train_data).double()
model = train(train_data)
filled_data = mida(model, test_data)
from RMSE import rmse
from pandas import DataFrame
filled_data = DataFrame(filled_data,
index=compare_data.index,
columns=compare_data.columns)
for index, col in enumerate(compare_data):
filled_data[col] = filled_data[col].astype(str(
compare_data[col].dtype))
err = rmse(filled_data, test_data)
print(filled_data, test_data)
print("err:", err)
def test_rmse4 (self) :
v = rmse(['5', '3', '2', '4', '5', '1', '4', '1', '5', '1', '5', '1', '2', '4', '5', '1', '3', '4', '2', '4'],
['2', '1', '3', '1', '1', '2', '4', '3', '1', '2', '2', '3', '2', '1', '1', '3', '4', '1', '5', '3'])
self.assert_(str(v) == '2.47991935353')
def test_rmse3 (self) :
v = rmse(['5', '3', '2', '4', '5'], ['2', '4', '3', '1', '2'])
self.assert_(str(v) == '2.40831891576')
def test_rmse2 (self) :
v = rmse(['1', '1', '1', '1', '1'], ['5', '5', '5', '5', '5'])
self.assert_(v == 4.0)
def test_rmse (self) :
v = rmse(['3', '3', '3'], ['3', '3', '3'])
self.assert_(v == 0.0)
