def featurize():
zero_dict_map = generate_zero_dict_map()
full_data = imdb.load_json('parsed.json')
train = filter(lambda x: (x['year'] < 2009), full_data)
test = filter(lambda x: (x['year'] >= 2009), full_data)
ERM2009 = imdb.load_json('ERM2009.json')
ERM_AVG = median(list((zip(*ERM2009.items()))[1]))
print 'ERM_MED', ERM_AVG
def dvectorize(kind, data):
zdict = deepcopy(zero_dict_map[kind])
if type(data) is not list: data = [data]
for feature in data:
if feature in zdict: zdict[feature] = 1
return zdict
def apply_featurization(movie):
for case in ['genres', 'mpaa', 'month']:
case_dict = dvectorize(case, movie[case])
for key in case_dict.keys():
movie[key] = case_dict[key]
for j in range(len(movie['stars'])):
actor_name = movie['stars'][j]['name']
if actor_name in ERM2009:
movie['actor' + str(j)] = ERM2009[actor_name]
else:
movie['actor' + str(j)] = ERM_AVG
for e_field in E_FIELDS:
VALS = []
for entity in movie[e_field]:
if entity['name'] in ERM2009:
VALS.append(ERM2009[entity['name']])
else:
VALS.append(ERM_AVG)
movie[e_field] = avg_float_list(VALS)
for thing in DELETE_THESE:
del movie[thing]
return movie
featurized_train = map(apply_featurization, train)
featurized_test = map(apply_featurization, test)
with open('featurized4_train.json', 'w') as f:
json.dump(featurized_train, f, sort_keys=True, indent=4)
with open('featurized4_test.json', 'w') as f:
json.dump(featurized_test, f, sort_keys=True, indent=4)
def featurize():
zero_dict_map = generate_zero_dict_map()
full_data = imdb.load_json('parsed.json')
train = filter(lambda x: (x['year'] < 2009), full_data)
test = filter(lambda x: (x['year'] >= 2009), full_data)
ERM2009 = imdb.load_json('ERM2009.json')
ERM_AVG = median(list((zip(*ERM2009.items()))[1]))
print 'ERM_MED', ERM_AVG
def dvectorize(kind, data):
zdict = deepcopy(zero_dict_map[kind])
if type(data) is not list: data = [data]
for feature in data:
if feature in zdict: zdict[feature] = 1
return zdict
def apply_featurization(movie):
for case in ['genres', 'mpaa', 'month']:
case_dict = dvectorize(case, movie[case])
for key in case_dict.keys():
movie[key] = case_dict[key]
for j in range(len(movie['stars'])):
actor_name = movie['stars'][j]['name']
if actor_name in ERM2009:
movie['actor'+str(j)] = ERM2009[actor_name]
else: movie['actor'+str(j)] = ERM_AVG
for e_field in E_FIELDS:
VALS = []
for entity in movie[e_field]:
if entity['name'] in ERM2009:
VALS.append(ERM2009[entity['name']])
else: VALS.append(ERM_AVG)
movie[e_field] = avg_float_list(VALS)
for thing in DELETE_THESE: del movie[thing]
return movie
featurized_train = map(apply_featurization, train)
featurized_test = map(apply_featurization, test)
with open('featurized4_train.json', 'w') as f:
json.dump(featurized_train, f, sort_keys = True, indent = 4)
with open('featurized4_test.json', 'w') as f:
json.dump(featurized_test, f, sort_keys = True, indent = 4)
def filter_data():
data = imdb.load_json('master.json')
newdata = sorted(data, key=lambda k: k['year'])
newdata.pop() # get rid of last movie with no fields
def assign_year(movie): # parse string year
year = int(str(movie['year'][1:5]))
movie['year'] = year # process year field
return movie
filtered = sorted(filter(my_filter, map(assign_year, newdata)), key=lambda k: k['year'])
print 'filtered length of %d' % len(filtered)
return filtered
def filter_data():
data = imdb.load_json('master.json')
newdata = sorted(data, key=lambda k: k['year'])
newdata.pop() # get rid of last movie with no fields
def assign_year(movie): # parse string year
year = int(str(movie['year'][1:5]))
movie['year'] = year # process year field
return movie
filtered = sorted(filter(my_filter, map(assign_year, newdata)), key=lambda k: k['year'])
print 'filtered length of %d' % len(filtered)
return filtered
def make_nice_graph(cls):
gdata = imdb.load_json('classify_results.json')
for key in gdata.keys():
model = key.split('(')[0]
scores = gdata[key][cls]
plt.plot(range(len(scores)), scores, label=model, linewidth=2)
if '0' in cls:
title = '< 100M Revenue (Class Label 0)'
else: title = '100M+ Revenue (Class Label 1)'
plt.legend()
plt.suptitle(title)
plt.xlabel('Time Step')
plt.ylabel('Classification Accuracy')
plt.show()
def make_nice_graph(cls):
gdata = imdb.load_json('classify_results.json')
for key in gdata.keys():
model = key.split('(')[0]
scores = gdata[key][cls]
plt.plot(range(len(scores)), scores, label=model, linewidth=2)
if '0' in cls:
title = '< 100M Revenue (Class Label 0)'
else:
title = '100M+ Revenue (Class Label 1)'
plt.legend()
plt.suptitle(title)
plt.xlabel('Time Step')
plt.ylabel('Classification Accuracy')
plt.show()
import imdb, json, math
#from textblob import TextBlob
from sklearn.svm import SVC
import matplotlib.pyplot as plt
from copy import deepcopy
from featurize import avg_float_list, MONTH_LIST
from sklearn.ensemble import RandomForestClassifier
from sklearn.naive_bayes import MultinomialNB, BernoulliNB
from sklearn.linear_model import LinearRegression, LogisticRegression, Ridge, Lasso
data = sorted(imdb.load_json('parsed.json'), key=lambda k: k['ordinaldate'])
fields = ['stars', 'directors', 'writers', 'production']
kmap = { 'month' : MONTH_LIST,
'mpaa' : ['G', 'PG_13', 'R', 'PG'],
'genres' : imdb.load_json('genres.json') }
M100, M50 = 100000000, 50000000
def labelbinarize(data):
imap = {}
for i in range(len(data)):
imap[data[i]] = i
return imap
def getLBMap():
LBMap = {}
for key in kmap.keys():
LBMap[key] = labelbinarize(kmap[key])
return LBMap
def generate_zero_dict_map():
return { 'genres' : zero_dict_from_list(imdb.load_json('genres.json')),
'month' : zero_dict_from_list(MONTH_LIST),
'mpaa' : zero_dict_from_list(['G', 'PG_13', 'R', 'PG']) }
import imdb, json, math
#from textblob import TextBlob
from sklearn.svm import SVC
import matplotlib.pyplot as plt
from copy import deepcopy
from featurize import avg_float_list, MONTH_LIST
from sklearn.ensemble import RandomForestClassifier
from sklearn.naive_bayes import MultinomialNB, BernoulliNB
from sklearn.linear_model import LinearRegression, LogisticRegression, Ridge, Lasso
data = sorted(imdb.load_json('parsed.json'), key=lambda k: k['ordinaldate'])
fields = ['stars', 'directors', 'writers', 'production']
kmap = {
'month': MONTH_LIST,
'mpaa': ['G', 'PG_13', 'R', 'PG'],
'genres': imdb.load_json('genres.json')
}
M100, M50 = 100000000, 50000000
def labelbinarize(data):
imap = {}
for i in range(len(data)):
imap[data[i]] = i
return imap
def getLBMap():
