A Recommendation System for Food Tourism

Dattatraya S. Adane1[0000−0003−3182−3328] , Himanshu

2[0009−0002−8173−1480]
Shahu , Parshva Choradia3[0009−0001−8704−5599] , and Ritesh
Yadav4[0009−0000−6881−2855]
1,2,3,4
Shri Ramdeobaba College of Engineering and Management, Nagpur MH
440013, INDIA
[email protected], [email protected], [email protected],
[email protected],

Abstract. By making personalised restaurant recommendations based

on visitors interests and needs, the Restaurant Recommender for Food
Tourism project seeks to improve tourists gastronomic experiences. A
trustworthy and effective system is required to help travellers choose the
best restaurants that suit their tastes and dietary requirements due to
the rising popularity of culinary tourism, which encourages visitors to
sample local cuisines and dining venues. In order to build a strong rec-
ommendation system, this project uses machine learning methods and
user data analysis. The recommender system creates personalised rec-
ommendations that match the person’s specific interests by considering
variables like cuisine type, price range, location, review counts and user
reviews. The system also has a user-friendly interface that enables users
to enter preferences with ease.

Keywords: Location-based recommendations, Local cuisines, Tourist

preferences

1 Introduction
Food tourism has become a significant trend in the travel industry, with an in-
creasing number of travellers seeking unique culinary experiences during their
trips. Exploring local cuisines and dining establishments has become an essential
part of the travel itinerary for many tourists. However, with an overwhelming
number of restaurants available in popular tourist destinations, it can be chal-
lenging for visitors to make informed decisions about where to dine. To address
this challenge, a recommendation system for food tourism can play a vital role.
Such a system leverages the power of technology, specifically machine learning
algorithms and user data analysis, to provide personalized restaurant recom-
mendations based on the preferences and requirements of individual tourists. By
considering factors such as cuisine type, price range, location, and user reviews,
the recommendation system assists tourists in selecting the best restaurants that
align with their specific tastes and dietary preferences. By implementing a re-
liable and efficient recommendation system, the project aims to enhance the
overall food tourism experience for travellers, facilitating their exploration of
local culinary cultures and supporting the growth of local restaurants.
2 Dattatraya et al.

2 Related Work
An overview of the literature on the subject is provided in this section. Sev-
eral research works have been carried out to extract domain-specific terms. A
comprehensive study of various Literature is:

In [1] Samad, Sabiha et al. focus on food consumption tracking and recom-
mendation apps available in major app stores. They created a rating tool to
assess app features and collected user feedback to understand user perspectives.
The review found that the apps had a number of drawbacks, such as inaccurate
ingredient predictions and a lack of automated food computation. Users’ ability
to locate foods unique to a particular region was hampered by the finite nature
of the databases used. Furthermore, dietitians’ expertise and the use of evidence-
based tactics were lacking in the majority of apps. The evaluation found that
current apps, which heavily rely on manual data entry, did not fully satisfy all
prerequisites for food tracking and recommendations. In addition to evidence-
based strategies and professional input, the researchers emphasized the need for
automatic food recognition and portion size estimates.

In [2] Martinez, Luis et al. demonstrate the integration of content-based fil-

tering and geographic distance radius in their recommender system, which has
produced encouraging results in resolving the cold-start issue. Users’ experiences
are enhanced and recommendations are more relevant when they take into ac-
count how close they are to restaurants. In addition, content-based filtering eval-
uates restaurant traits and characteristics to produce tailored recommendations
based on user preferences. This combination enables our system to deliver pre-
cise and location-aware recommendations even to first-time users. The outcomes
show improved user satisfaction, better recommendation quality, and successful
cold-start problem mitigation. Their recommender system successfully addresses
the issues related to new user recommendations by utilizing geographic distance
radius and content-based filtering.

In [3] Krishnaraj, Sathya Seelan et al. focuses on a recommendation system

that utilizes various algorithms to suggest products to customers based on histor-
ical data. The system is divided into two categories: collaborative filtering, which
recognizes patterns in user behaviour to make recommendations, and content-
based filtering, which makes suggestions for items based on their characteristics
and user preferences. The latter proved to be the most successful for rating pre-
diction between users and restaurants after they experimented with Slope One,
k-Nearest Neighbors, and multiclass SVM classification. They combined collab-
orative and popularity-based filtering, as well as using content-based filtering, to
address the cold start issue. Their analysis revealed increased user-friendliness
and efficiency of restaurant prediction. They successfully solved the cold start
issue without the use of other methods by utilizing content-based filtering.
In [4] Munaji, Achmad concentrated on creating a restaurant recommenda-
tion system using collaborative filtering algorithms based on user ratings. Users
 A Recommendation System for Food Tourism 3

with similar preferences and interests are assumed to have strong correlations in
their choices, which allows them to recommend products to one another based
on their prior preferences. But they found that collaborative filtering had a
number of shortcomings. Insufficient user data or the addition of new users or
items to the system can cause the ”cold start problem.” Due to the lack of
data on which to base recommendations, accuracy becomes difficult in such cir-
cumstances. Additionally, data sparsity is a common problem in collaborative
filtering because it needs a significant amount of user-item interaction data to
produce accurate predictions. Unfortunately, the availability of such informa-
tion can be limited, which can reduce the accuracy of recommendations. The
researchers compared collaborative and content-based filtering for recommenda-
tions in order to overcome these limitations. The cold-start issue is reduced by
content-based filtering, which makes suggestions based on the properties and
features of the items. By taking into account each user’s preferences and unique
characteristics, content-based filtering also improves personalization. Overall,
the study demonstrates how collaborative filtering-related issues can be resolved
and recommendation systems can be improved using content-based filtering. By
incorporating content-based techniques, recommendation systems can give users
more precise and customized suggestions, enhancing the user experience and
boosting system adoption.

In [5] Asani addresses the utilization of recommender systems to assist users

in making informed decisions across various domains, including products and ser-
vices. The authors highlight the importance of recommender systems in aiding
tourists and travellers in choosing restaurants among numerous and unfamiliar
options. Extracting user preferences in recommender systems is a complex task.
Conventional methods like surveys and user ratings aren’t always reliable. In
today’s digital age, online comments on platforms like websites and social media
offer a wealth of implicit information. To tackle this, the authors suggest harness-
ing natural language processing to extract food preferences from these comments
by analyzing the sentiments expressed. The authors introduce a method to im-
prove user preference extraction using sentiment analysis. Their recommender
system extracts preferences from comments, uses text mining and noun extrac-
tion, filters with WordNet, clusters nouns semantically, and applies sentiment
analysis to score clusters. The top-scored cluster reflects user preferences. Simi-
larity between preferences and restaurant menus guides recommendations.

In [6] Blanca Vargas-Govea et al, Integrated contextual data into recom-

mender system for increasing user satisfaction. As it is customary to include a
broad variety of contextual variables, but owing to possible dimensionality issues,
it is crucial to choose the most important onces. For finding relevant contextual
data they studied and used selection methods. They consider this feature se-
lection as a classification challenge and they used it in restaurant recommender
system that integrates semantic insights and user annotations. There outcomes
show that a smaller collection of qualities may be used to maintain or enhance
system performance, showing the potential of feature selection methods.
4 Dattatraya et al.

3 Proposed Methodology

3.1 Data Collection

The First step we took in developing a recommendation system was to find reli-
able data sources that provide information on restaurants, menus, user reviews,
and ratings. For the given problems, we web scraped the data of restaurants and
local vendors of google map using Beautiful Soup. We collected data of around
2100 restaurants having its name, address, rating, no. of people rated, contact
information, address and many more. The extracted data is saved and organized
in a structured format as a CSV file.

Figure 3.1: CSV File of Raw Data

3.2 Data Preprocessing

The next step was Data cleaning and preparation for our food tourism recom-
mendation system. So, we cleaned the data by removing the NaN values from
the rating and rating count sections and duplicate values of restaurant. We also
dropped the unnecessary columns that has no role in recommendation. To en-
sure consistency, we standardized the data formats, and through geocoding we
made it possible to calculate accurate distances for recommendations based on
proximity. By incorporating these preprocessing stages, meal suggestions gain
authority and accuracy, giving users individualized, top-notch culinary tourism
experiences.
 A Recommendation System for Food Tourism 5

Figure 3.2: Preprocessed Data

3.3 Content-Based Filtering

Content-based filtering uses item features to recommend other items similar to

what the user likes, based on their previous actions or explicit feedback. So our
next step was Content-based filtering, we used it in food tourism recommenda-
tion systems to match user preferences with food items. We use attributes like
food category, cuisine type and restaurant location. By leveraging content-based
filtering, the system offers personalized and relevant food options that cater to
the user’s specific preferences, enhancing their food tourism experience.

3.4 User-Preferences and location

In the system, we take the user’s interests and location into account to provide
personalized recommendations. Users have the option to specify food preferences.
The system generates relevant recommendations based on the user’s current
location and preferences.

3.5 Interface and Presentation

To increase the number of customers, we made the user interface of the appli-
cation simple and easy to use. To enable the quick entry of selections, we used
drop-down menus. Recommendations are visually appealing and include infor-
mation about the suggested restaurants, their locations in relation to the user,
6 Dattatraya et al.

and ratings and reviews of the restaurants. Making decisions is aided by addi-
tional information such as menus and restaurant descriptions. For a satisfying
culinary tourism experience, an interface with a straightforward structure that
encourages discovery and offers a wealth of interesting information is used. In
the figure below, the UI for the home page, user input, and a drop-down menu
for choosing a category and distance are all displayed.

Figure 3.3: Home Page UI

3.6 Integration of Website with the Backend

We integrated a website into a backend involves seamlessly connecting the fron-

tend, developed with HTML, CSS, and JavaScript, with a backend server like
Flask. HTML structures the content, CSS enhances the design, and JavaScript
adds interactivity. Flask, a Python web framework, handles requests and re-
sponses, processing data and logic behind the scenes. This integration enables
dynamic web applications, where user interactions trigger server actions, provid-
ing a smooth and responsive user experience.

3.7 Taking Input from the User and Displaying the Output

We used drop-down menus to allow customers to select food categories and took
their current location. We allowed customers to specify the restaurant’s distance
requirements by using the Distance Radius Range functionality. We validate and
sanitize user input for accuracy and to prevent processing issues. Connecting to
the backend will allow us to send user preferences for the creation of recommen-
dations. We provide recommendations that take distance and food category into
account based on user preferences. The results of the recommendations were sent
back to the website in a formatted form.
 A Recommendation System for Food Tourism 7

Figure 3.4 : Taking User Input

Figure 3.5 : Displaying Output

4 Result and Evaluation

We built a Website using HTML, CSS, and Javascript and used Beautiful Soup
for web scraping. We collected comprehensive and reliable food-related infor-
mation, such as restaurant listings, menus, user reviews, and ratings. This data
served as the foundation for the recommendation system. Implementing content-
based filtering allows the system to match user preferences with food items based
on their content attributes, such as food category, cuisine type, location, and dis-
tance. By considering the user’s inputs for food category, location, and range of
the restaurant, the system can generate personalized recommendations that align
with the user’s preferences. The results of the food tourism recommender sys-
tem would ideally provide the user with a list of recommended restaurants that
match their specified food category and are within the desired range from their
location. The recommendations should consider the content attributes and pri-
oritize restaurants that closely align with the user’s preferences. The system can
8 Dattatraya et al.

present the recommendations in the website’s user interface, displaying relevant

information such as restaurant names, cuisines, ratings, and distances.

Figure 4.1: User Interface

Figure 4.2: Taking User Preferences

 A Recommendation System for Food Tourism 9

Figure 4.3: Submitting the Preferences

Figure 4.4: Displaying the Output

In Figure 4.1, The User Interface of the Website is shown. The user inter-
face for the restaurant recommendation system features two drop-down menus
that allow the user to input their preferences. The first input is titled ”Distance
Radius (in Km)” and offers options for the user to choose the desired radius
within which they want to find restaurants. This user input includes a variety
of distance ranges, such as 1 km, 5 km, 10 km, 20 km, etc. The user can select
the ideal radius of distance based on their location or preferences. As seen in
Figure 4.2, there is a drop-down menu with the title ”Select the Category” or
”Open this to Select Menu” that enables users to choose the kind of restau-
rant they are looking for. This drop-down menu offers a range of categories,
such as North Indian Food, Chinese, Vegetarian Restaurants, Indian, Fast Food,
Bakery, and more. The user can select the desired category from the available
options. The restaurant recommendation process can be started by clicking a
button or submitting a form after the user has made their selections in both
user input fields, as shown in Figure 4.3. A list of suggested restaurants that fit
the chosen distance radius and category will then be generated by the system
after processing the user’s inputs. A Submit button to start the recommenda-
tion process and a display area where the suggested restaurants are shown are
10 Dattatraya et al.

features of the user interface that were added to improve the user experience.
According to Figure 4.4, the display area can show pertinent information about
each restaurant, including its name, address, ratings, and URL. By selecting the
radius of travel and the type of restaurant, users can specify their preferences
in a simple and intuitive manner through the use of this user interface design.
It makes it possible for them to get personalized restaurant recommendations
that fit their needs, enhancing their dining experience and making it simpler for
them to locate suitable dining options.

References
1. Samad, S. A., Naher, F., Kabir, S., Das, A., Amin, A., Islam, S.
S., Mohammed, S. (2022). Smartphone Apps for Tracking Food Con-
sumption and Recommendations: Evaluating Artificial Intelligence-based
Functionalities, Features and Quality of Current Apps. ResearchGate.
https://www.researchgate.net/publication/362489544 Smartphone Apps for
Tracking Food Consumption and Recommendations Evaluating Artificial
Intelligence-based Functionalities Features and Quality of Current Apps
2. Martı́nez, L. R., Espinilla, M. (2009). REJA: A Geo-referenced Hy-
brid Recommender System for Restaurants. Web Intelligence and Intelligent
Agent Technology, IEEE/WIC/ACM International Conference on, 3, 187-190.
10.1109/WI-IAT.2009.259. https://www.researchgate.net/publication/232625008
REJA A Georeferenced Hybrid Recommender System for Restaurants
3. Krishnaraj, S. S. (2021). Restaurant Recommendation System Using Machine
Learning Algorithms. RIET-IJSET International Journal of Science Engineer-
ing and Technology, 7, 1-8. https://www.researchgate.net/publication/354238686
Restaurant Recommendation System Using Machine Learning Algorithms
4. Munaji, A. E. A. (2021). Restaurant Recommendation System Based on User
Ratings with Collaborative Filtering. IOP Conference Series: Materials Sci-
ence and Engineering, 1077, 012026. 10.1088/1757-899X/1077/1/012026. .https:
//www.researchgate.net/publication/350084914 Restaurant Recommendation
System Based on User Ratings with Collaborative Filtering
5. Asani, E. V. N., Sadri, H., Javad. (2021). Restaurant Recommender Sys-
tem Based on Sentiment Analysis. Machine Learning with Applications, 6,
100114. 10.1016/j.mlwa.2021.100114. https://www.researchgate.net/publication/
353429372 Restaurant recommender system based on sentiment analysis
6. Vargas Govea, B. S., Ponce-Medell ı́n, G., Rafael. (2011). Effects of Relevant
Contextual Features in the Performance of a Restaurant Recommender System.
791. https://www.researchgate.net/publication/248703303 Effects of relevant
contextual features in the performance of a restaurant recommender system