def setup_libraries(self):
# Usage:
# The request object is an HTTPResponse object in Django. It is able to be used
# like an dictionary to get http parameters, such as ?num_samples=X&facebook_Group=Y.
# We will try to get num_samples, facebook_group, and pusher_uuid. Otherwise
# we will raise a SuspiciousOperation if it fails. Afterwards get data from the facebook
# group, and then train our model, then use our model to determine sentiment of users.
# Arguments:
# yelp_extractor (YelpExtractor) : a yelp data extractor that can extract yelp data
# facebook_extractor (FacebookExtractor) : a facebook extractor that can extract facebook data
# logistic_classifier (LogisticRegression) : a logistic regression class that can compute sentiment of a user
# Return:
# None
# Setup our Libraries
self.yelp_extractor = YelpExtractor()
self.facebook_extractor = FacebookExtractor()
self.logistic_classifier = LogisticClassifier()
def setup_libraries(self):
# Usage:
# The request object is an HTTPResponse object in Django. It is able to be used
# like an dictionary to get http parameters, such as ?num_samples=X&facebook_Group=Y.
# We will try to get num_samples, facebook_group, and pusher_uuid. Otherwise
# we will raise a SuspiciousOperation if it fails. Afterwards get data from the facebook
# group, and then train our model, then use our model to determine sentiment of users.
# Arguments:
# yelp_extractor (YelpExtractor) : a yelp data extractor that can extract yelp data
# facebook_extractor (FacebookExtractor) : a facebook extractor that can extract facebook data
# logistic_classifier (LogisticRegression) : a logistic regression class that can compute sentiment of a user
# Return:
# None
# Setup our Libraries
self.yelp_extractor = YelpExtractor()
self.facebook_extractor = FacebookExtractor()
self.logistic_classifier = LogisticClassifier()
class Train(View):
# Usage:
# This is used to train a SFrame logistic regression model. And return the data using a pusher.
# The URL parameters should contain num_samples, facebook group name, and pusher UUID channel
# We will get the group data such as members from the facebook group name, and train the
# logistic regression model with num_samples of data from amazon, and yelp (each half of num_samples)
def setup_libraries(self):
# Usage:
# The request object is an HTTPResponse object in Django. It is able to be used
# like an dictionary to get http parameters, such as ?num_samples=X&facebook_Group=Y.
# We will try to get num_samples, facebook_group, and pusher_uuid. Otherwise
# we will raise a SuspiciousOperation if it fails. Afterwards get data from the facebook
# group, and then train our model, then use our model to determine sentiment of users.
# Arguments:
# yelp_extractor (YelpExtractor) : a yelp data extractor that can extract yelp data
# facebook_extractor (FacebookExtractor) : a facebook extractor that can extract facebook data
# logistic_classifier (LogisticRegression) : a logistic regression class that can compute sentiment of a user
# Return:
# None
# Setup our Libraries
self.yelp_extractor = YelpExtractor()
self.facebook_extractor = FacebookExtractor()
self.logistic_classifier = LogisticClassifier()
def get(self, request):
# Usage:
# The request object is an HTTPResponse object in Django. It is able to be used
# like an dictionary to get http parameters, such as ?num_samples=X&facebook_Group=Y.
# We will try to get num_samples, facebook_group, and pusher_uuid. Otherwise
# we will raise a SuspiciousOperation if it fails. Afterwards get data from the facebook
# group, and then train our model, then use our model to determine sentiment of users.
# Arguments:
# request (object) : django request object
# Return:
# Json (JsonResponse) : a Json object with status = done
# Setup our libraries first
self.setup_libraries()
# get our url parameters
num_samples, facebook_group = self.get_url_parameters(request)
# use the group to create models for our facebook group
facebook_message, facebook_count = self.get_facebook_group(facebook_group=facebook_group)
# get our yelp reviews appended into a database, and we can get some message, and count
yelp_message, yelp_exception, yelp_count = self.get_yelp_reviews(num_samples=math.ceil(num_samples/2))
# set user_sentiment to None
user_sentiment = None
# set status to success, and message to None
status = "Success"
message = None
# Need to test whether if we got any yelp data
if YelpReview.objects.count() > 0:
# train, and get our data for user's sentiment
user_sentiment = self.compute_sentiment_analysis(model_iterators=[YelpReview.objects.all().iterator()],
num_samples=num_samples,
facebook_group=facebook_group,
facebook_user_model=FacebookUser,
facebook_post_model=FacebookPost)
else:
# There is no reviews, hence we will set a message
status = "Failure"
message = "There are no reviews"
# Return a JsonResponse to update to the front-end that the status is success
return JsonResponse({'status':status, 'message':message, 'yelp_message':yelp_message, 'yelp_exception':yelp_exception,
'yelp_count':yelp_count,'facebook_message':facebook_message, 'facebook_count':facebook_count,
'user_sentiment':user_sentiment})
def get_url_parameters(self, request):
# Usage:
# The request object is an HTTPResponse object in Django. It is able to be used
# like an dictionary to get http parameters, such as ?num_samples=X&facebook_Group=Y.
# We will try to get num_samples, facebook_group, and pusher_uuid. Otherwise
# we will return a HttpResponseBadRequest if it fails.
# Arguments:
# request (object) : django request object
# Return:
# num_samples (int) : number of samples to train with
# facebook_group (string) : facebook_group string, can be a number or a string
# try to get a few url parameters, such as num_samples, facebook_group, pusher_uuid
try:
# num_samples should be an integer, and facebook_group are strings
num_samples = int(request.GET.get('num_samples'))
facebook_group = request.GET.get('facebook_group')
# return the tuples of the url parameters
return num_samples, facebook_group
except KeyError:
# If we fail to get any of the url parameters, we will raise an exception for Invalid url parameters
raise SuspiciousOperation('Invalid url parameters')
except Exception as e:
# Fails due to unknown reasons
raise SuspiciousOperation('Unknown Error:%s' % str(e))
def get_facebook_group(self, facebook_group):
# Usage:
# This function will utilize the facebook_extractor to get the data using the
# facebook graph api, and append the data into the facebook model.
# Arguments:
# facebook_group (string) : name of the facebook group
# Return:
# message (string) : Message of the current status
# count (int) : Number of users extracted from yelp
# First, get or create a facebook_group
facebook_group_model, facebook_group_model_created = FacebookGroup.objects.get_or_create(id=facebook_group,
defaults=dict(id=facebook_group))
# Loop through the messages using the facebook_extractor
for data in self.facebook_extractor.distribute_get_posts(group=facebook_group):
# Get or create a facebook user, since that user might exist in other public groups as well
facebook_user, facebook_user_created = FacebookUser.objects.get_or_create(id=data["user_id"],
defaults=dict(username=data["user_name"],
facebook_group=facebook_group_model))
# Update or create the user's messages
FacebookPost.objects.update_or_create(id=data["message_id"], post=data["message_text"],
facebook_user=facebook_user, defaults=dict(id=data["message_id"],
post=data["message_text"],
facebook_user=facebook_user))
# Return a facebook extractor status, and the number of users in the group
return "Successfully finished", FacebookUser.objects.filter(facebook_group=facebook_group).count()
def get_yelp_reviews(self, num_samples):
# Usage:
# This function will utilize the yelp_extractor to get the data using the
# yelp api, and append the data into the yelp model.
# Arguments:
# num_samples (int) : number of samples to get from yelp
# Return:
# message (string) : Message of the current status
# count (int) : Number of samples extracted from yelp
# Get how many counts there are for yelp reviews before adding
yelp_reviews_before = YelpReview.objects.count()
# If there are less than reviews we needed, then we will add more into the database
if yelp_reviews_before < num_samples:
# distributed_find_reviews is a generator that will return reviews as a dictionary format
# of dict(data, exception)
for review in self.yelp_extractor.distributed_find_reviews(reviews_required=num_samples-yelp_reviews_before):
# If there is an issue raised inside the generator
if review["exception"]:
# Return the exception message, and how many count was added
return review["exception"].message, review["exception"].errors, abs(YelpReview.objects.count()-yelp_reviews_before)
# Add a new YelpReview into the database
YelpReview(id=review["data"]["id"], review_text=review["data"]["review"], star=review["data"]["star"]).save()
# There was enough reviews in the first place
else:
# Return a message saying that extraction was not required, and the new reviews added was 0
return "No need for extraction, already have enough reviews", None, 0
# Return a success message, and how many count was added
return "Yelp extractor completed successfully", None, abs(YelpReview.objects.count()-yelp_reviews_before)
def compute_sentiment_analysis(self, model_iterators, num_samples, facebook_group, facebook_user_model, facebook_post_model):
# Usage:
# This function will utilize the logistic classifier class to
# compute sentiment analysis
# Arguments:
# model_iterators (list) : list of model ierators
# each model should have ID, Review, and Star
# num_samples (int) : the number of samples to train. These will be
# divided amongst the model_iterators.
# facebook_group (string) : facebook group number or ID
# facebook_user_model (model) : facebook user model
# facebook_post_model (model) : facebook post model
# Return:
# array of dictionary (array) : an array with dictionary user_id, mean_predicted_sentiment
# Add models to the logistic classifier
self.logistic_classifier.add_models(model_iterators, num_samples)
# Train the models with the current models
self.logistic_classifier.train_model()
# Add Facebook model data to the logistic classifier
self.logistic_classifier.add_facebook_model(facebook_group, facebook_user_model, facebook_post_model)
# Predict the sentiment for all the facebook posts
self.logistic_classifier.predict_sentiment()
# Average out each user's sentiment
self.logistic_classifier.average_user_sentiment()
return [dict(user_id=sentiment["user_id"],
username=sentiment["username"],
mean_predicted_sentiment=sentiment["mean_predicted_sentiment"])
for sentiment in self.logistic_classifier.get_sentiment()]
class Train(View):
# Usage:
# This is used to train a SFrame logistic regression model. And return the data using a pusher.
# The URL parameters should contain num_samples, facebook group name, and pusher UUID channel
# We will get the group data such as members from the facebook group name, and train the
# logistic regression model with num_samples of data from amazon, and yelp (each half of num_samples)
def setup_libraries(self):
# Usage:
# The request object is an HTTPResponse object in Django. It is able to be used
# like an dictionary to get http parameters, such as ?num_samples=X&facebook_Group=Y.
# We will try to get num_samples, facebook_group, and pusher_uuid. Otherwise
# we will raise a SuspiciousOperation if it fails. Afterwards get data from the facebook
# group, and then train our model, then use our model to determine sentiment of users.
# Arguments:
# yelp_extractor (YelpExtractor) : a yelp data extractor that can extract yelp data
# facebook_extractor (FacebookExtractor) : a facebook extractor that can extract facebook data
# logistic_classifier (LogisticRegression) : a logistic regression class that can compute sentiment of a user
# Return:
# None
# Setup our Libraries
self.yelp_extractor = YelpExtractor()
self.facebook_extractor = FacebookExtractor()
self.logistic_classifier = LogisticClassifier()
def get(self, request):
# Usage:
# The request object is an HTTPResponse object in Django. It is able to be used
# like an dictionary to get http parameters, such as ?num_samples=X&facebook_Group=Y.
# We will try to get num_samples, facebook_group, and pusher_uuid. Otherwise
# we will raise a SuspiciousOperation if it fails. Afterwards get data from the facebook
# group, and then train our model, then use our model to determine sentiment of users.
# Arguments:
# request (object) : django request object
# Return:
# Json (JsonResponse) : a Json object with status = done
# Setup our libraries first
self.setup_libraries()
# get our url parameters
num_samples, facebook_group = self.get_url_parameters(request)
# use the group to create models for our facebook group
facebook_message, facebook_count = self.get_facebook_group(
facebook_group=facebook_group)
# get our yelp reviews appended into a database, and we can get some message, and count
yelp_message, yelp_exception, yelp_count = self.get_yelp_reviews(
num_samples=math.ceil(num_samples / 2))
# set user_sentiment to None
user_sentiment = None
# set status to success, and message to None
status = "Success"
message = None
# Need to test whether if we got any yelp data
if YelpReview.objects.count() > 0:
# train, and get our data for user's sentiment
user_sentiment = self.compute_sentiment_analysis(
model_iterators=[YelpReview.objects.all().iterator()],
num_samples=num_samples,
facebook_group=facebook_group,
facebook_user_model=FacebookUser,
facebook_post_model=FacebookPost)
else:
# There is no reviews, hence we will set a message
status = "Failure"
message = "There are no reviews"
# Return a JsonResponse to update to the front-end that the status is success
return JsonResponse({
'status': status,
'message': message,
'yelp_message': yelp_message,
'yelp_exception': yelp_exception,
'yelp_count': yelp_count,
'facebook_message': facebook_message,
'facebook_count': facebook_count,
'user_sentiment': user_sentiment
})
def get_url_parameters(self, request):
# Usage:
# The request object is an HTTPResponse object in Django. It is able to be used
# like an dictionary to get http parameters, such as ?num_samples=X&facebook_Group=Y.
# We will try to get num_samples, facebook_group, and pusher_uuid. Otherwise
# we will return a HttpResponseBadRequest if it fails.
# Arguments:
# request (object) : django request object
# Return:
# num_samples (int) : number of samples to train with
# facebook_group (string) : facebook_group string, can be a number or a string
# try to get a few url parameters, such as num_samples, facebook_group, pusher_uuid
try:
# num_samples should be an integer, and facebook_group are strings
num_samples = int(request.GET.get('num_samples'))
facebook_group = request.GET.get('facebook_group')
# return the tuples of the url parameters
return num_samples, facebook_group
except KeyError:
# If we fail to get any of the url parameters, we will raise an exception for Invalid url parameters
raise SuspiciousOperation('Invalid url parameters')
except Exception as e:
# Fails due to unknown reasons
raise SuspiciousOperation('Unknown Error:%s' % str(e))
def get_facebook_group(self, facebook_group):
# Usage:
# This function will utilize the facebook_extractor to get the data using the
# facebook graph api, and append the data into the facebook model.
# Arguments:
# facebook_group (string) : name of the facebook group
# Return:
# message (string) : Message of the current status
# count (int) : Number of users extracted from yelp
# First, get or create a facebook_group
facebook_group_model, facebook_group_model_created = FacebookGroup.objects.get_or_create(
id=facebook_group, defaults=dict(id=facebook_group))
# Loop through the messages using the facebook_extractor
for data in self.facebook_extractor.distribute_get_posts(
group=facebook_group):
# Get or create a facebook user, since that user might exist in other public groups as well
facebook_user, facebook_user_created = FacebookUser.objects.get_or_create(
id=data["user_id"],
defaults=dict(username=data["user_name"],
facebook_group=facebook_group_model))
# Update or create the user's messages
FacebookPost.objects.update_or_create(
id=data["message_id"],
post=data["message_text"],
facebook_user=facebook_user,
defaults=dict(id=data["message_id"],
post=data["message_text"],
facebook_user=facebook_user))
# Return a facebook extractor status, and the number of users in the group
return "Successfully finished", FacebookUser.objects.filter(
facebook_group=facebook_group).count()
def get_yelp_reviews(self, num_samples):
# Usage:
# This function will utilize the yelp_extractor to get the data using the
# yelp api, and append the data into the yelp model.
# Arguments:
# num_samples (int) : number of samples to get from yelp
# Return:
# message (string) : Message of the current status
# count (int) : Number of samples extracted from yelp
# Get how many counts there are for yelp reviews before adding
yelp_reviews_before = YelpReview.objects.count()
# If there are less than reviews we needed, then we will add more into the database
if yelp_reviews_before < num_samples:
# distributed_find_reviews is a generator that will return reviews as a dictionary format
# of dict(data, exception)
for review in self.yelp_extractor.distributed_find_reviews(
reviews_required=num_samples - yelp_reviews_before):
# If there is an issue raised inside the generator
if review["exception"]:
# Return the exception message, and how many count was added
return review["exception"].message, review[
"exception"].errors, abs(YelpReview.objects.count() -
yelp_reviews_before)
# Add a new YelpReview into the database
YelpReview(id=review["data"]["id"],
review_text=review["data"]["review"],
star=review["data"]["star"]).save()
# There was enough reviews in the first place
else:
# Return a message saying that extraction was not required, and the new reviews added was 0
return "No need for extraction, already have enough reviews", None, 0
# Return a success message, and how many count was added
return "Yelp extractor completed successfully", None, abs(
YelpReview.objects.count() - yelp_reviews_before)
def compute_sentiment_analysis(self, model_iterators, num_samples,
facebook_group, facebook_user_model,
facebook_post_model):
# Usage:
# This function will utilize the logistic classifier class to
# compute sentiment analysis
# Arguments:
# model_iterators (list) : list of model ierators
# each model should have ID, Review, and Star
# num_samples (int) : the number of samples to train. These will be
# divided amongst the model_iterators.
# facebook_group (string) : facebook group number or ID
# facebook_user_model (model) : facebook user model
# facebook_post_model (model) : facebook post model
# Return:
# array of dictionary (array) : an array with dictionary user_id, mean_predicted_sentiment
# Add models to the logistic classifier
self.logistic_classifier.add_models(model_iterators, num_samples)
# Train the models with the current models
self.logistic_classifier.train_model()
# Add Facebook model data to the logistic classifier
self.logistic_classifier.add_facebook_model(facebook_group,
facebook_user_model,
facebook_post_model)
# Predict the sentiment for all the facebook posts
self.logistic_classifier.predict_sentiment()
# Average out each user's sentiment
self.logistic_classifier.average_user_sentiment()
return [
dict(
user_id=sentiment["user_id"],
username=sentiment["username"],
mean_predicted_sentiment=sentiment["mean_predicted_sentiment"])
for sentiment in self.logistic_classifier.get_sentiment()
]
