This blog post discusses how to quickly save a trained machine learning model that suggests recipes based on the user’s data, and nutritional value using Joblib and Pickle. The reason I decided to work on this is that the domain of food is varied and complex and presents many challenges to developing a recommender system. In addition to the nutritional value, price, preparation time, and cultural aspects of a meal, ingredients and cooking methods play a significant role. Diet and nutrition are complex domains for recommenders that are required to assist users with starting and maintaining healthy lifestyles. This post is a continuation from here.

The first recommender uses publicly available data from Kaggle and mock data for the user.

The steps involved in the project are as follows:

Quickly Save trained machine learning model using Joblib and Pickle. The model is able to predict calorie levels from the user’s demographic, preferences, and nutritional value.

2. Use the predicted calorie level to randomly suggest a recipe from the database.

3. Build and deploy a web application

ADataAnalyst is reader-supported. As an Amazon Associate, I earn from qualifying purchases when you buy through links on our site

Save Trained Machine Learning Model

EDA_21042021_Blog

In [1]:

import numpy as np # linear algebra
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
from sklearn.ensemble import RandomForestClassifier
pd.set_option('display.max_columns', None)
pd.set_option('display.max_rows', None)

In [2]:

pp_recipes = pd.read_csv("archive/PP_recipes.csv")
pp_users = pd.read_csv("archive/PP_users.csv")
raw_interactions = pd.read_csv("archive/RAW_interactions.csv")
raw_recipes = pd.read_csv("archive/RAW_recipes.csv")

In [3]:

pp_recipes.head(2)

Out[3]:

	id	i	name_tokens	ingredient_tokens	steps_tokens	techniques	calorie_level	ingredient_ids
0	424415	23	[40480, 37229, 2911, 1019, 249, 6878, 6878, 28…	[[2911, 1019, 249, 6878], [1353], [6953], [153…	[40480, 40482, 21662, 481, 6878, 500, 246, 161…	[0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, …	0	[389, 7655, 6270, 1527, 3406]
1	146223	96900	[40480, 18376, 7056, 246, 1531, 2032, 40481]	[[17918], [25916], [2507, 6444], [8467, 1179],…	[40480, 40482, 729, 2525, 10906, 485, 43, 8393…	[1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, …	0	[2683, 4969, 800, 5298, 840, 2499, 6632, 7022,…

In [4]:

pp_users.head(2)

Out[4]:

	u	techniques	items	n_items	ratings	n_ratings
0	0	[8, 0, 0, 5, 6, 0, 0, 1, 0, 9, 1, 0, 0, 0, 1, …	[1118, 27680, 32541, 137353, 16428, 28815, 658…	31	[5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 4.0, 4.0, …	31
1	1	[11, 0, 0, 2, 12, 0, 0, 0, 0, 14, 5, 0, 0, 0, …	[122140, 77036, 156817, 76957, 68818, 155600, …	39	[5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, …	39

In [5]:

raw_interactions.head(2)

Out[5]:

	user_id	recipe_id	date	rating	review
0	38094	40893	2003-02-17	4	Great with a salad. Cooked on top of stove for…
1	1293707	40893	2011-12-21	5	So simple, so delicious! Great for chilly fall…

In [6]:

raw_recipes.head(2)

Out[6]:

	name	id	minutes	contributor_id	submitted	tags	nutrition	n_steps	steps	description	ingredients	n_ingredients
0	arriba baked winter squash mexican style	137739	55	47892	2005-09-16	[’60-minutes-or-less’, ‘time-to-make’, ‘course…	[51.5, 0.0, 13.0, 0.0, 2.0, 0.0, 4.0]	11	[‘make a choice and proceed with recipe’, ‘dep…	autumn is my favorite time of year to cook! th…	[‘winter squash’, ‘mexican seasoning’, ‘mixed …	7
1	a bit different breakfast pizza	31490	30	26278	2002-06-17	[’30-minutes-or-less’, ‘time-to-make’, ‘course…	[173.4, 18.0, 0.0, 17.0, 22.0, 35.0, 1.0]	9	[‘preheat oven to 425 degrees f’, ‘press dough…	this recipe calls for the crust to be prebaked…	[‘prepared pizza crust’, ‘sausage patty’, ‘egg…	6

In [7]:

df = raw_interactions.merge(raw_recipes, left_on = 'recipe_id', right_on= 'id')

In [8]:

df = df.merge(pp_users, left_on = 'user_id', right_on= 'u').merge(pp_recipes, left_on = 'recipe_id', right_on= 'i')

In [9]:

df = df[[ 'u','recipe_id','rating', 'review', 'name',
       'minutes', 'tags', 'nutrition',
       'n_steps', 'steps', 'description', 'ingredients', 'n_ingredients',  'calorie_level']]

In [13]:

mock = pd.read_csv('archive/mock_data.csv')
mock1 = pd.read_csv('archive/mock_data1.csv')
mock2 = pd.read_csv('archive/mock_data2.csv')
mock3 = pd.read_csv('archive/mock_data3.csv')
mock4 = pd.read_csv('archive/mock_data4.csv')
mock5 = pd.read_csv('archive/mock_data5.csv')
mock6 = pd.read_csv('archive/mock_data6.csv')
mock7 = pd.read_csv('archive/mock_data7.csv')

all_mock = pd.concat([mock, mock1, mock2, mock3, mock4, mock5, mock6, mock7], sort=False)

In [14]:

all_mock = all_mock.drop_duplicates('u')

In [15]:

all_mock.head(2)

Out[15]:

	u	gender	age	prefer_food	id	race
0	24780.0	F	50	Squash – Sunburst	NaN	NaN
1	2728.0	F	69	Glucose	NaN	NaN

In [ ]:

df = df.merge(all_mock, on ='u')

In [ ]:

df[['calories','total fat (PDV)','sugar (PDV)','sodium (PDV)','protein (PDV)','saturated fat (PDV)','carbohydrates (PDV)']] = df.nutrition.str.split(",",expand=True)

In [ ]:

df['calories'] =  df['calories'].apply(lambda x: x.replace('[',''))

In [ ]:

df['carbohydrates (PDV)'] =  df['carbohydrates (PDV)'].apply(lambda x: x.replace(']',''))

In [ ]:

df[['calories','total fat (PDV)','sugar (PDV)','sodium (PDV)','protein (PDV)','saturated fat (PDV)','carbohydrates (PDV)']] = df[['calories','total fat (PDV)','sugar (PDV)','sodium (PDV)','protein (PDV)','saturated fat (PDV)','carbohydrates (PDV)']].astype('float')

In [ ]:

df['food types'] = np.nan
df['food types'] = df['food types'].astype('str')

In [ ]:

for i in df['ingredients'].index:
    if('eggs' not in df['ingredients'][i]):
         if('ice-cream' in df['ingredients'][i] or 'chocolate' in df['ingredients'][i] or 'cookies' in df['ingredients'][i]):
                df['food types'][i]='Veg dessert'
    elif('eggs' in df['ingredients'][i]):
        if('ice-cream' in df['ingredients'][i] or 'chocolate' in df['ingredients'][i] or 'cookies' in df['ingredients'][i]):
                df['food types'][i]='Non-Veg dessert'

In [ ]:

for i in df.index:
    if(df['food types'][i]!='Veg dessert' and df['food types'][i]!='Non-Veg dessert' and 20<df['calories'][i]<300):
        df['food types'][i]='Healthy'

In [ ]:

for i in df.index:
    if(df['food types'][i]!='Veg dessert' and df['food types'][i]!='Non-Veg dessert' and df['food types'][i]!='Healthy'):
        if('chicken' in df['ingredients'][i] or 'eggs' in df['ingredients'][i] or'ham' in df['ingredients'][i] or 'pepperoni' in df['ingredients'][i] ):
            df['food types'][i]='Non-veg'

In [ ]:

for i in df.index:
    if(df['food types'][i]!='Veg dessert' and df['food types'][i]!='Non-Veg dessert' and df['food types'][i]!='Healthy' and df['food types'][i]!='Non-veg'):
        df['food types'][i]='Veg'

In [ ]:

df.rename(columns={"food types": "food_types"}, inplace = True)

In [ ]:

df['food_types'] = df['food_types'].replace({'Healthy' : 'Low-Calories'})

In [ ]:

def create_diff(row):
    if row < 10:
        return 'Easy'
    elif 10 < row < 20:
        return 'Medium'
    else:
        return 'Hard'

df['difficulty'] = df['n_steps'].apply(lambda x : create_diff(x))

In [ ]:

model_df = df[[ 'gender', 'age', 'calories','difficulty','food_types', 'calorie_level']]

In [ ]:

X = model_df.drop('calorie_level', axis = 1)
y = model_df['calorie_level']

In [ ]:

gender_dict = {"F": 0, "M": 1}
difficulty_dict = {'Easy': 0, 'Medium': 1, 'Hard': 2}
food_dict = {'Low-Calories' : 0, 'Non-veg' : 1, 'Veg' : 2, 'Veg dessert': 3, 'Non-Veg dessert': 4}
X.gender.replace(gender_dict, inplace = True)
X.difficulty.replace(difficulty_dict, inplace = True)
X.food_types.replace(food_dict, inplace = True)

In [ ]:

clf = RandomForestClassifier(max_depth=2, random_state=0)
clf.fit(X, y)

In [ ]:

import pickle
model_path =  "model.pkl"
pickle.dump(clf, open(model_path, 'wb'))

In [ ]:

recipe_df = df[[ 'calorie_level','food_types','name','description', 'ingredients', 'minutes', 'steps', 'total fat (PDV)', 'sugar (PDV)', 'sodium (PDV)',
       'protein (PDV)', 'saturated fat (PDV)', 'carbohydrates (PDV)',
       ]]
recipe_df.food_types.replace(food_dict, inplace = True)

In [ ]:

recipe_df.to_csv("recipes.csv", index = False)

In the following post, I am going to create a frontend to the pickle file of the model and deploy it using:

GitHub Repository
Streamlit Sharing
AWS EC2 instance
AWS Beanstalk
AWS Fargate

I am also going to assess the advantages and disadvantages of each deployment technique.

References

https://www.kaggle.com/aayushmishra1512/food-recommender
https://towardsdatascience.com/building-a-food-recommendation-system-90788f78691a
https://www.kaggle.com/prmohanty/python-how-to-save-and-load-ml-models

Tagged Machine Learning, MLOps, Modelling

How to Quickly Save a Trained Machine Learning Model

Save Trained Machine Learning Model

References