IRJET- Recipe Recommendation System using Machine Learning Models
0 likes110 views
This document describes a recipe recommendation system that uses machine learning models to suggest ingredient pairs and alternative ingredients. It discusses using a vector space model and Word2Vec model to find highly similar ingredient pairs from different cuisines and recommend substitutes. The models are trained on recipe data scraped from online databases. The system aims to help users innovate new dishes and accommodate allergies. It outlines collecting recipe data, preprocessing it, calculating ingredient similarities, and using the models to suggest pairings and alternatives.
![International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 09 | Sep 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 366
Recipe Recommendation System using Machine Learning Models
Suyash Maheshwari1, Manas Chourey2
1Student (B.E.), Information Technology, Sinhgad College Of Engineering, Pune, India
2Student (B.E.), Information Technology, Sinhgad College Of Engineering, Pune, India
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract - Completing recipes is a challenging task, as the
success of ingredient combinations depends on a multitude of
factors such as taste, smell and texture. We can pair various
ingredients with each other and generate a dish. However, itis
not possible to perform these experiments in kitchen. Hence,
science is used to determine various ingredient pairs. Due to
the increasing availability of data in online databases, data
mining and machine learning methods are starting to play a
prominent role in food consumption analysis and food
preference modeling. Here, we will use two machine learning
models-vector space model and Word2Vec model to find top
ingredient pairs from different cuisines and to suggest
alternate ingredients. The focusisontheIndiancuisine. Indian
cuisine is very vast and diverse and hence it is difficult to find
patterns and generate pairs.
Key Words: Data analysis, Machine learning, Web
scraping.
1. INTRODUCTION
Everyday food is an important part of our daily lives and,
currently, there are many internet sites that help us plan
these meals. Allied to the growth in the amount of food data
such as recipes available on the internet there is an increase
in the number of studies on these data, such as recipe
analysis and recipe search. However, there are few publicly
available resources for food research especially for the
Indian cuisine; those that do exist do not include a wide
range of food data or any meal data that is, likely
combinations of recipes. In this project, we will construct a
food recommendation system using big data analytics.
1.1 Aim And Objectives
The aim is to develop a tool that would help in pairing
various ingredients from different cuisinesandtosuggest an
alternate ingredient.
The objective is to help in innovation of new dishes and to
help people allergic to certain ingredientsbyrecommending
alternate ingredients.
2. LITERATURE SURVEY
2.1 Data-driven recipe completion using machine
learning methods
In this paper Marlies De Clercq, Michiel Stock, Bernard De
Baets and Willem Waegeman [1] have proposed that
machine learning methods can be used to build models to
complete a recipe. The machine learning model suggested is
Non negative matrix factorization. The characteristic of this
method is that it gives a linear non negative part based
representation of data. But, it does not consider the
information about flavor components and hence the results
are not accurate.
2.2 Alternative-ingredient Recommendation Based on
Co-occurrence Relation on Recipe Database
Ryosuke Yamanishia, Naoki Shinoa,YokoNishiharaa,Junichi
Fukumotoa and Aya Kaizakib[2] have proposed a
recommendation methodofalternativeingredients basedon
co-occurrence relation on recipe database. This paper
proposes two algorithms to recommend alternative
ingredient. Through the cooking and tasting experiments, it
was confirmed that the each of the proposed methods were
effective for each intended purpose. Through cooking
demonstration withthe recommendedalternativeingredient
and subjective evaluationexperiments,it wasconfirmed that
the both algorithm recommended acceptableingredientsfor
over 90 percent. However, algorithm 1 might not
recommend alternative-ingredient similar to the exchange
ingredient.
3. METHODOLOGY
3.1 Web Scraping
Web Scraping is a technique employed to extract large
amounts of data from websiteswherebythedata is extracted
and saved to a local file in computer or to a database in table
format. Web scraping is used in order to extract the data
from Yummly.com for later retrieval and analysis. Yummly
is a website that provides recipe recommendations
personalized to the individual's taste, semantic recipe
search. The data is taken from the Yummly websites which
contains various recipes, ingredients and also the flavour
components of all the ingredients. This data is then stored
into the csv file for further operations. Pkl files are also
created. A PKL file is pickled to savespacewhenbeingstored
or transferred over a network and is unpickled and loaded
back into program memory during runtime. After the data is
scraped, it is cleaned and preprocessedtoremoveanomalies
from the data.
3.2 Ingredient Pairing
The ingredient pairing module isusedtopairtheingredients
of the selected recipe with other ingredients to innovate a
new dish. For this, Tf-idf and Cosine similarity are used. Tf-
idf (Term frequency - Inverse document frequency) assigns
weights to ingredients based on their flavor components.](https://image.slidesharecdn.com/irjet-v6i946-191206093434/85/IRJET-Recipe-Recommendation-System-using-Machine-Learning-Models-1-320.jpg)
![International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 09 | Sep 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 368
6. CONCLUSIONS
Food recommendation system includes food pairing and
alternative ingredient recommendation.Foodpairing isused
to pair the ingredients to help the restaurants or other users
to innovate new dishes. Tf-idf and cosinesimilaritymethods
are used for food pairing. Tf-Idf i.e. Term frequency-Inverse
document frequency is used to search theingredientstopair
with other ingredients. Cosine similarity is used to find
cosine angle between two vectors and the vectors having
least angle between them are recommended as ingredients.
Alternative IngredientRecommendationsystemisstudiedto
replace an ingredient in the recipe and recommend an
alternative ingredient. Word2vec model is used to find the
similar ingredients and the most similar ingredients are
recommended with their similarity value to that of the
ingredient that is to be replaced. The proposed system can
be used by users to innovate new dishes and replace any
ingredient in the recipe to complete the recipe.
6.1 End User Dashboard
Fig -4: Home Page
Fig -5: Select ingredient for pairing
Fig -6: Paired ingredients
Fig -7: Select ingredient to find alternative
Fig -8: Alternate ingredients
REFERENCES
[1] Marlies De Clercq, Michiel Stock, Bernard De Baets,
Willem Waegeman, "Data-driven recipe completion using
machine learning methods". Trends in food science and
technology, ISSN 0924-2244, Volume: 49, 1-13, 2016](https://image.slidesharecdn.com/irjet-v6i946-191206093434/85/IRJET-Recipe-Recommendation-System-using-Machine-Learning-Models-3-320.jpg)
![International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 09 | Sep 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 369
[2] Ryosuke Yamanishia, Naoki Shinoa, Yoko Nishiharaa,
Junichi Fukumotoa, Aya Kaizakib, "Alternative ingredient
Recommendation Based on Co-occurrence Relation on
Recipe Database". Procedia Computer Science 60 (2015)
986-993
[3] Jain, Anupam, N. K. Rakhi, and Ganesh Bagler. ”Analysis
of food pairing in regional cuisines of India.” PloS one
10.10 (2015): e0139539.](https://image.slidesharecdn.com/irjet-v6i946-191206093434/85/IRJET-Recipe-Recommendation-System-using-Machine-Learning-Models-4-320.jpg)
IRJET- Recipe Recommendation System using Machine Learning Models
- 1. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 06 Issue: 09 | Sep 2019 www.irjet.net p-ISSN: 2395-0072 © 2019, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 366 Recipe Recommendation System using Machine Learning Models Suyash Maheshwari1, Manas Chourey2 1Student (B.E.), Information Technology, Sinhgad College Of Engineering, Pune, India 2Student (B.E.), Information Technology, Sinhgad College Of Engineering, Pune, India ---------------------------------------------------------------------***---------------------------------------------------------------------- Abstract - Completing recipes is a challenging task, as the success of ingredient combinations depends on a multitude of factors such as taste, smell and texture. We can pair various ingredients with each other and generate a dish. However, itis not possible to perform these experiments in kitchen. Hence, science is used to determine various ingredient pairs. Due to the increasing availability of data in online databases, data mining and machine learning methods are starting to play a prominent role in food consumption analysis and food preference modeling. Here, we will use two machine learning models-vector space model and Word2Vec model to find top ingredient pairs from different cuisines and to suggest alternate ingredients. The focusisontheIndiancuisine. Indian cuisine is very vast and diverse and hence it is difficult to find patterns and generate pairs. Key Words: Data analysis, Machine learning, Web scraping. 1. INTRODUCTION Everyday food is an important part of our daily lives and, currently, there are many internet sites that help us plan these meals. Allied to the growth in the amount of food data such as recipes available on the internet there is an increase in the number of studies on these data, such as recipe analysis and recipe search. However, there are few publicly available resources for food research especially for the Indian cuisine; those that do exist do not include a wide range of food data or any meal data that is, likely combinations of recipes. In this project, we will construct a food recommendation system using big data analytics. 1.1 Aim And Objectives The aim is to develop a tool that would help in pairing various ingredients from different cuisinesandtosuggest an alternate ingredient. The objective is to help in innovation of new dishes and to help people allergic to certain ingredientsbyrecommending alternate ingredients. 2. LITERATURE SURVEY 2.1 Data-driven recipe completion using machine learning methods In this paper Marlies De Clercq, Michiel Stock, Bernard De Baets and Willem Waegeman [1] have proposed that machine learning methods can be used to build models to complete a recipe. The machine learning model suggested is Non negative matrix factorization. The characteristic of this method is that it gives a linear non negative part based representation of data. But, it does not consider the information about flavor components and hence the results are not accurate. 2.2 Alternative-ingredient Recommendation Based on Co-occurrence Relation on Recipe Database Ryosuke Yamanishia, Naoki Shinoa,YokoNishiharaa,Junichi Fukumotoa and Aya Kaizakib[2] have proposed a recommendation methodofalternativeingredients basedon co-occurrence relation on recipe database. This paper proposes two algorithms to recommend alternative ingredient. Through the cooking and tasting experiments, it was confirmed that the each of the proposed methods were effective for each intended purpose. Through cooking demonstration withthe recommendedalternativeingredient and subjective evaluationexperiments,it wasconfirmed that the both algorithm recommended acceptableingredientsfor over 90 percent. However, algorithm 1 might not recommend alternative-ingredient similar to the exchange ingredient. 3. METHODOLOGY 3.1 Web Scraping Web Scraping is a technique employed to extract large amounts of data from websiteswherebythedata is extracted and saved to a local file in computer or to a database in table format. Web scraping is used in order to extract the data from Yummly.com for later retrieval and analysis. Yummly is a website that provides recipe recommendations personalized to the individual's taste, semantic recipe search. The data is taken from the Yummly websites which contains various recipes, ingredients and also the flavour components of all the ingredients. This data is then stored into the csv file for further operations. Pkl files are also created. A PKL file is pickled to savespacewhenbeingstored or transferred over a network and is unpickled and loaded back into program memory during runtime. After the data is scraped, it is cleaned and preprocessedtoremoveanomalies from the data. 3.2 Ingredient Pairing The ingredient pairing module isusedtopairtheingredients of the selected recipe with other ingredients to innovate a new dish. For this, Tf-idf and Cosine similarity are used. Tf- idf (Term frequency - Inverse document frequency) assigns weights to ingredients based on their flavor components.
- 2. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 06 Issue: 09 | Sep 2019 www.irjet.net p-ISSN: 2395-0072 © 2019, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 367 These Tf-idf scores are then used to calculate cosine similarity for two ingredients. The model used for calculation of cosine similarity is known as vector space model. Based on the similarity, different ingredients from various cuisines are recommended for pairing. Fig -1: Ingredient Pairing A matrix is formed for all ingredients present in dataset and all the flavor components. Each ingredient is treated as a single document and flavor componentsaretreatedasterms of documents. Tf-idf score is calculated for all ingredients on the basis of this matrix. Then, for the selected ingredient , matching pair is found by calculating the cosine of the angle between two ingredients. The lesser the angle,the betterthe pairs are. This model is defined as Vector space model. 3.2 Alternate Ingredient Recommendation Alternative Ingredient recommendation is used to recommend an alternative ingredient if a particular ingredient is not present or cannot be used in the recipe. Word2vec model is created to perform this function and recommend alternative ingredients. Word2vec takes as its input as large list of ingredients andproducesa vector space, typically of several hundred dimensions, with each unique ingredient in the corpus being assigned a corresponding vector in the space. Vectors are positioned in the vector space such that ingredients that share common contexts in the corpus are located in close proximity to one another in the space. Gensim library is used to automatically extract semantic topics from documents, as efficiently as possible. Downsampling is also done to help reducing data size. At last, most similar() method is used to find top-nmostsimilar ingredients. This method computes cosine similarity between a simple mean of the projection weight vectors of the given words and the vectors for each word in the model. This method returns the sequence of ingredients and their similarity value to the ingredient that is to be exchanged. 4. FOOD PAIRING HYPOTHESIS Food pairing hypothesis states that two ingredients sharing common flavor profile taste well when used together in a recipe. In the Indian cuisine vegetables, fruits, cereals, pulses, nuts support the food pairing hypothesis whereas spices and dairy products do not support this hypothesis as shown in the figure below. An analysis was done to find out the relationship between ingredients and their flavor profiles. It showed a linear dependency. Fig -2: Analysis of Indian Cuisine 5. STATE DIAGRAM A state diagram is used to represent the condition of the system or part of the system at finite instances of time. It’s a behavioral diagram and it represents the behavior using finite state transitions. A state diagram is used to model the dynamic behavior of a classinresponsetotimeandchanging external stimuli. Fig -3: state diagram
- 3. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 06 Issue: 09 | Sep 2019 www.irjet.net p-ISSN: 2395-0072 © 2019, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 368 6. CONCLUSIONS Food recommendation system includes food pairing and alternative ingredient recommendation.Foodpairing isused to pair the ingredients to help the restaurants or other users to innovate new dishes. Tf-idf and cosinesimilaritymethods are used for food pairing. Tf-Idf i.e. Term frequency-Inverse document frequency is used to search theingredientstopair with other ingredients. Cosine similarity is used to find cosine angle between two vectors and the vectors having least angle between them are recommended as ingredients. Alternative IngredientRecommendationsystemisstudiedto replace an ingredient in the recipe and recommend an alternative ingredient. Word2vec model is used to find the similar ingredients and the most similar ingredients are recommended with their similarity value to that of the ingredient that is to be replaced. The proposed system can be used by users to innovate new dishes and replace any ingredient in the recipe to complete the recipe. 6.1 End User Dashboard Fig -4: Home Page Fig -5: Select ingredient for pairing Fig -6: Paired ingredients Fig -7: Select ingredient to find alternative Fig -8: Alternate ingredients REFERENCES [1] Marlies De Clercq, Michiel Stock, Bernard De Baets, Willem Waegeman, "Data-driven recipe completion using machine learning methods". Trends in food science and technology, ISSN 0924-2244, Volume: 49, 1-13, 2016
- 4. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 06 Issue: 09 | Sep 2019 www.irjet.net p-ISSN: 2395-0072 © 2019, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 369 [2] Ryosuke Yamanishia, Naoki Shinoa, Yoko Nishiharaa, Junichi Fukumotoa, Aya Kaizakib, "Alternative ingredient Recommendation Based on Co-occurrence Relation on Recipe Database". Procedia Computer Science 60 (2015) 986-993 [3] Jain, Anupam, N. K. Rakhi, and Ganesh Bagler. ”Analysis of food pairing in regional cuisines of India.” PloS one 10.10 (2015): e0139539.