def trainModel(mtype, dataset):
# first column of dataset contains ids
# the next 10 columns contain the feature vector
array = dataset.values
I = array[:, 0]
# if it is a supervised model the 2nd column is the expected result
if (mtype > 3):
R = array[:, 1]
X = array[:2:]
validation_size = 0.20
seed = 7
X_train, X_validation, Y_train, Y_validation = \
cross_validation.train_test_split(X, Y, test_size=validation_size,
random_state=seed)
num_folds = 10
num_instances = len(X_train)
scoring = 'accuracy'
else:
X = array[:, 1:]
if (mtype == 1): # do unsupervised Kmeans
kmeans = KMeans(n_clusters=2, random_state=0).fit(X)
data.updateResult(list(zip(kmeans.labels_, I)))
model = kmeans
elif (mtype == 3): # supervised NB
nb = GaussianNB()
nb.fit(X_train, Y_train)
model = nb
def setTModel(mtype):
(reviews, names) = data.getTrainData()
dataset = pandas.DataFrame(data = reviews, columns=names)
print ("num train samples " + str(dataset.shape))
# first column of dataset contains ids
# the next 10 columns contain the feature vector
array=dataset.values
I = array[:,0]
# in a supervised model the 2nd column is the expected result, else,
# column where the result will be populated
R = array[:,1]
X = array[:,2:]
validation_size = 0.20
seed = 7
X_train, X_validation, Y_train, Y_validation = \
cross_validation.train_test_split(X, R, test_size=validation_size,
random_state=seed)
num_folds = 10
num_instances = len(X_train)
scoring = 'accuracy'
if (mtype == 1): # do unsupervised Kmeans
tmodel = KMeans(n_clusters = 2, random_state = 0).fit(X)
data.updateResult(list(zip(tmodel.labels_, I)))
data.commit()
elif (mtype == 3): # supervised NB
tmodel = GaussianNB()
tmodel.fit(X_train, Y_train)
else:
tmodel = ()
return tmodel
def create_product():
if not request.get_json(force=True) or not 'asin' in request.get_json(force=True):
abort(400)
prod_id = request.get_json(force=True)['asin']
cat_id = request.get_json(force=True)['cid']
pname = request.get_json(force=True)['title']
url = request.get_json(force=True)['imurl']
price = request.get_json(force=True)['price']
rank = request.get_json(force=True)['rank']
if len(prod_id) == 0:
abort(404)
# need to check if product is in the product table, if not create it, else throw error
if data.existProduct(prod_id):
abort(405)
product = {'pid': prod_id, 'cid': cat_id, 'pname': pname, 'imurl': url, 'price': price, 'rank': rank}
data.insertProductFeatures(product)
reviews = request.get_json(force=True)['reviews']
for r in reviews:
ret = requests.post(sa_url, data=json.dumps({'review': r, 'product': product}))
review = json.loads(ret.text)
if data.existReview(prod_id, review['rid']):
abort(405)
# review with that prod_id and reviewer_id should not exist, but if they
# do throw error
print(review)
data.commit()
(uneval_reviews, names) = data.getUnevalReviews(prod_id)
print('found some uneval reviews')
if uneval_reviews is not None:
print(json.dumps({'reviews': uneval_reviews, 'names': names}))
r = requests.post(ml_url, data=json.dumps({'reviews': uneval_reviews, 'names': names}))
result = json.loads(r.text)['result']
data.updateResult(result)
product = data.getProductInfo(prod_id)
return jsonify({'product': product[0]})
def get_product(prod_id):
print "here?"
(uneval_reviews, names) = data.getUnevalReviews(prod_id)
if uneval_reviews is not None:
print('found some uneval reviews')
print(json.dumps({'reviews': uneval_reviews, 'names': names}))
r = requests.post(ml_url, data=json.dumps({'reviews': uneval_reviews, 'names': names}))
result = json.loads(r.text)['result']
data.updateResult(result)
product = data.getProductInfo(prod_id)
if len(product) == 0:
abort(404)
return jsonify({'product': product[0]})
I = array[:, 0] # in a supervised model the 2nd column is the expected result, else, # column where the result will be populated R = array[:, 1] X = array[:, 2:] #validation_size = 0.20 #seed = 7 #X_train, X_validation, Y_train, Y_validation = \ # cross_validation.train_test_split(X, R, test_size=validation_size, # random_state=seed) #num_folds = 10 #num_instances = len(X_train) #scoring = 'accuracy' kmeans = KMeans(n_clusters=2, random_state=1).fit(X) data.updateResult(zip(I, kmeans.labels_)) (reviews, names) = data.getUnTrainData() dataset = pandas.DataFrame(data=reviews, columns=names) print(dataset.shape) # first column of dataset contains ids # the next 10 columns contain the feature vector array = dataset.values I = array[:, 0] X = array[:, 2:] data.updateResult(zip(I, kmeans.predict(X))) (reviews, names) = data.getAllReviewFeatures() dataset = pandas.DataFrame(data=reviews, columns=names)