def train(Id):
Id = Id
content = get_data_sql(Id)
#param = content.get('parameteres')
algo = content.get('algorithmes')
req_data = request.json
path_db = req_data['path_db'] #lib/data/train.csv"
cdir = os.path.dirname(__file__)
file_res = os.path.join(cdir, path_db)
data = pd.read_csv(file_res)
y = data[data.columns[-1]]
X = data.drop(data.columns[-1], axis=1)
if not os.path.exists('models'):
os.mkdir('models')
nom = algo + str(Id)
score, X, y = build_model(True, X, y, 0.3, nom, algo)
mydb = connect_to_mysql()
path = 'models/' + nom + '.pkl'
mycursor = mydb.cursor()
sql = "UPDATE MODELS SET path='" + path + "' where id=" + str(Id)
mycursor.execute(sql)
mydb.commit()
post_data = {}
post_data['Success'] = 'True'
post_data['precision'] = str(score)
return str(post_data)
def predict(Id):
Id = Id
req_data = request.json
path_db = req_data['path_db']
content = get_data_sql(Id)
#split = content.get('parameteres')
path_pkl = content.get('path')
algo = content.get('algorithmes')
cdir = os.path.dirname(__file__)
file_res = os.path.join(cdir, path_db)
data = pd.read_csv(file_res)
X = data
y = None
nom = algo + str(Id)
#On applique le pre-traitement
score, X, y = build_model(False, X, y, 0.2, nom, algo)
loaded_model = pickle.load(open(path_pkl, 'rb'))
pred = loaded_model.predict(X)
post_data = {}
post_data['Success'] = 'True'
post_data['prediction'] = pred
return str(post_data)
def split():
req_data = request.args
model_name = req_data['split']
print(model_name)
build_model(float(model_name))
return 'sucess'