def one_hot_form_input(brand, item_type, title, est_price):
"""
Returns pandas series
"""
X_sample, _ = joblib.load(reg_model_path)
one_hot_array = X_sample
one_hot_array['cost'] = est_price
if brand in one_hot_array.index.values:
one_hot_array[brand] = 1
else:
one_hot_array['other brand'] = 1
if item_type in one_hot_array.index.values:
one_hot_array[item_type] = 1
else:
pass
adjectives_query = "SELECT * FROM item_adjectives;"
adjectives = [a[1] for a in engine.execute(adjectives_query).fetchall()]
for adj in adjectives:
match = re.search('{}'.format(adj), title, re.IGNORECASE)
if match:
#l.append(1)
try:
one_hot_array[adj] = 1
except TypeError:
pass
# take the exponent because the model was trained on the log of rent values
# prediction = np.exp(reg.predict(input_df.T)[0])
return one_hot_array
def find_by_type_and_brand(item_type, brand):
item_type = item_type.replace("'", "''")
brand = brand.replace("'", "''")
query = "SELECT I.id, I.brand, I.item_type, I.cost, I.sku, \
I.rent_per_week, I.created_at, I.title, I.description, \
I.year_purchased, I.rent_per_week, \
R.rental_date, R.return_date, \
RI.item_price, RI.refunded, RI.fit_return, \
RI.created_at, \
RI.updated_at \
FROM items I \
LEFT JOIN rental_items RI ON I.id = RI.item_id \
LEFT JOIN rentals R ON R.id = RI.rental_id \
WHERE item_type='{}' and brand='{}';".format(item_type, brand)
return engine.execute(query).fetchall()
from wombat.engine import ml_model
from wombat.models import engine, dbsession, Item
from wtforms import Form, StringField, SubmitField, SelectField, FloatField, validators
item_types = engine.execute('SELECT DISTINCT item_type,\
count(item_type) FROM items GROUP BY item_type ORDER BY \
count(item_type) DESC;').fetchall()
item_types = [(r[0], r[0].title()) for r in item_types]
brands_query = "SELECT * FROM brands;"
brands = ml_model.brands[0:100]
brand_names = sorted([b[1] for b in engine.execute(brands_query).fetchall()],
key=str.lower)
brand_names.remove('other brand')
brands = [(b, b) for b in brand_names]
brands = [('', 'Choose a brand')] + [('other brand', 'Other')] + brands
class DescriptionForm(Form):
description = StringField('Description')
item_type = SelectField(
label='Category',
choices=item_types,
validators=[validators.Required(message='Category is required')])
brand = SelectField(
label='Brand',
choices=brands,
validators=[validators.Required(message='Please choose a brand')])
est_price = FloatField(
label='Retail Price',
validators=[
validators.Required(message="""Please enter a retail price. If you
if part[1] in speech_parts:
p_stemmer = PorterStemmer()
word = p_stemmer.stem(part[0].lower())
if (word not in existing_words) and (word not in l) and (word not in banned_words):
l.append("{}\n".format(word))
print("l is: {}".format(l))
print("word is: {}".format(word))
except Exception as e:
print(str(e))
print("l is: {}".format(l))
with open('key_words2.txt', 'a') as fp:
for word in l:
fp.write(word)
def get_first_sentence(string):
try:
return re.split(r'(?<=[.:;])\s', string)[0]
except TypeError:
return ''
res = engine.execute("SELECT title, description FROM items WHERE brand != 'LENDER SUBMISSION FILL IN' AND rent_per_week < 1000").fetchall()
for text in res[0:10]:
first_sentence = get_first_sentence(text[1])
item_title = text[0]
combined = ' '.join([item_title, first_sentence])
process_sentence(combined)
# print(res[0][0])
# process_sentence2(0)
# Usually we just select all the brands but sometimes we need to limit the
# number we choose for testing purposes and it makes sense just to pick the
# most popular
brands_query = "SELECT brand, count(brand) FROM items WHERE brand != 'LENDER SUBMISSION FILL IN' AND rent_per_week < {} GROUP BY brand ORDER BY count(brand) DESC;".format(
rent_per_week_max)
brand_df = pd.read_sql_query(brands_query, engine)
brands_escaped = [
"\'{}\'".format(brand.replace("'", "''")) for brand in brand_df['brand']
]
brands_escaped = ', '.join(brands_escaped)
# create the list of brands so other modules can access what brands are being
# used to create the model
res = engine.execute(brands_query).fetchall()
brands = [r[0] for r in res]
brand_length = len(brands)
# grab items form db to train model.
# training query is the canonical query that the machine learning model is
# based on. If you change it then you have to reconstuct the model
canonical_query = "SELECT brand, item_type, title, cost, rent_per_week, description FROM items WHERE brand in ({}) AND rent_per_week < {}".format(
brands_escaped, rent_per_week_max)
df = pd.read_sql_query(canonical_query, engine)
canonical_df = df
# get one-hot columns for brands
dummified_brands = pd.get_dummies(df['brand'])
df = pd.concat([df, dummified_brands], axis=1)
df = df.drop('brand', axis=1)
def get_brands():
res = engine.execute('SELECT DISTINCT brand FROM items;').fetchall()
return [brand[0] for brand in res]
def get_item_types():
res = engine.execute('SELECT DISTINCT item_type FROM items;').fetchall()
return [item[0] for item in res]