research paper no 1

FOOD RECOMMENDATION SYSTEM

Paper 1 develops a food recommender system that integrates nutritional information

and user preferences to generate personalized meal plans. The approach uses MCDA
for filtering and optimization scenarios for short-term intelligent models,
resulting in daily meal plans that address both nutritional needs and user
preferences. This fills a gap in managing both user preferences and nutritional
information in food recommendation systems.

Paper 2 proposes a food recommendation system for diabetic patients based on

nutrition and food characteristics. The approach uses SOM algorithm and K-mean
clustering to classify food groups considering eight significant nutrients, aiming
to improve food categorization for diabetes diet care beyond existing
ontology-based methods. This enhances diet management for diabetic patients through
better food categorization.

Paper 3 presents a context-aware food recommendation system that provides real-time

recommendations based on user profiles, physiological signals, and environmental
information. The approach utilizes the u-BabSang system to integrate individual
profiles, physiological status, and environmental context, enabling personalized
well-being care by considering various contextual factors in real-time. This
improves health monitoring and personalized recommendations by analyzing individual
circumstances.

Paper 4 reviews and classifies recent research on deep learning-based recommender

systems, offering a taxonomy of deep learning-based recommendation models and
summarizing state-of-the-art advancements. The paper highlights notable works,
discusses advantages and disadvantages of deep learning methods, and identifies
future research directions, providing a comprehensive understanding and framework
for future studies in deep learning-based recommendation systems.

Paper 5 systematically reviews the use of machine learning algorithms in

recommender systems, analyzing various ML algorithms, their characteristics, and
application in RS, and discussing the role of software engineering in RS
development. The paper identifies trends in ML algorithm usage, highlights
challenges, and suggests areas for future research, offering insights into ML
algorithm selection for RS and the integration of software engineering tools
in RS development.

METHODOLOGY
There are two primary phases in the system:
1. Training phase:
The framework is prepared by utilizing the information within the information set
and fits a show (line/curve)
based on the calculation chosen appropriately.
2. Testing phase:
The framework is given with the inputs and is tried for its working. The accuracy
is checked. And therefore, the
data that is used to train the model or test it, has to be [Link] system
is designed to detect the
accurate food for the people. and appropriate algorithm is used to do the task.
Before the algorithms were selected for further use, different algorithms were
compared for their accuracy. The
well-suited one for the task was chosen.

3.1. ALGORITHMS:
1. CONTENT FILTERING ALGORITHM
The content-based suggestion calculations determine from data recovery and sifting
inquire about. The
content-based suggestion is the continuation and advancement of early collaborative
filtering strategy. Content
based proposal systems suggest things comparable to those that the client has
chosen within the past instead of
the user’s comments of things.

K-NEAREST NEIGHBORS ALGORITHM:

KNN is a machine learning algorithm to find clusters of similar users based on and
make predictions using
cosine similarity. We can understand it’s working with the help of following steps:

I.
For actualizing any calculation, we require an information set. So amid the primary
step of KNN, we must
stack the preparing as well as test data.
Next, we have to select the esteem of K i.e. the closest information focuses. K can
be any integer.
For each point within the test data do the following –
i. Calculate the remove between test information and each push of preparing data
with the assistance of
cosine similarity.
ii. Now, based on the removed esteem, sort them in rising order.
iii. Next, it'll select the best K lines from the sorted array.
iv. Now, it'll relegate a lesson to the test point based on the foremost visit
lesson of these rows.

reference FOOD RECOMMENDATION SYSTEM

Atharva Parulekar*1, Juhi Vichare*2, Jay Tarkar*3, Prof. Prachi Kshirsagar*4
*1,2,3Students Of Information Technology Engineering, Padmabhushan Vasantdada Patil
Pratishthan’s
College Of Engineering, Maharashtra, India.
4Prof. Of Department Of Information Technology Engineering, Padmabhushan Vasantdada
Patil
Pratishthan’s College Of Engineering, Maharashtra, India.
research paper no 2

Implementing Customizable Online Food Ordering System

Using Web Based Application

**Full Service Restaurant**: The traditional food order process in full-service

restaurants involves a waiter bringing a paper-based menu to guests, who then
select items and inform the waiter. This requires guests to be seated and a waiter
to assist with ordering. The conventional paper-based system has drawbacks such as
the risk of losing or damaging records, wastage of money, time, and paper, and lack
of dynamicity. Changes to the menu require reprinting, and significant human effort
is required, making it time-consuming and prone to errors from the customer's
perspective.

**Self Service Restaurant**: In self-service restaurants, guests place orders at

the service counter. They must decide on their order before reaching the counter,
with the menu often presented as posters behind the counter. This system requires
guests to know what they want in advance, simplifying the ordering process but
limiting menu flexibility.

**Automated Food Ordering System**: Some restaurants use automation to reduce

service costs and enhance customer experiences. These systems typically involve
electronic devices with screens that display the menu and accept orders. Waiters
enter orders into a system, which displays the information in the kitchen. The
kitchen staff prepares the dishes, informs the waiter, and the waiter delivers the
food to the table. The system also tracks dish availability and allows for
modifications or cancellations if a dish is unavailable. Bills are generated based
on orders, and management has access to all customer details in the system.
Advances in computer and communication technology have led to the development of
various computerized food ordering systems.

The technologies which are used to implement the system are:

Visual Studio 2010 for developing web application.
Android version 2.2 or more for Tablets is required.
We have used Android 4.2 Jellybean as the working
platform to develop this application.
SQL 2008 is a light weight Database which is
going to be used for database access from the
tablet.

reference:-
Implementing Customizable Online Food Ordering System
Using Web Based Application
Varsha Chavan, Priya Jadhav, Snehal Korade and Priyanka Teli
Computer Department,Pune University
Indapur, Maharashtra,India

research paper no 3:

Restaurant Recommendation System Using

Machine Learning Algorithms

Various methods have been developed for creating restaurant recommendation systems,
each employing distinct approaches and technologies to enhance accuracy and user
satisfaction. One such approach focuses on user preferences, utilizing topic
modeling to uncover hidden aspects in review texts and employing regression models
to establish user-restaurant relationships. This system aims to improve
recommendations by understanding the multifaceted factors influencing user
preferences. Another system integrates personalized location-based recommendations
into mobile technology, studying user behavioral patterns to refine recommendation
accuracy.

Machine learning algorithms play a significant role in another approach, where

several collaborative filtering methods, including Slope One, k-Nearest Neighbors,
and multiclass SVM classification, are tested for predicting ratings. Among these,
the multiclass SVM method outperformed others, demonstrating its effectiveness in
handling complex recommendation tasks. Additionally, user-based and item-based
collaborative filtering algorithms were compared for rating predictions, leading to
the development of an architecture supporting real-time recommendation services.

A proposed system uses Support Vector Machine (SVM) to predict restaurants based on
user location. This system aims to save users time, effort, and money by
recommending restaurants considering factors like cuisine type, location, ambiance,
price range, popularity, and ratings. Data from platforms like Yelp and Zomato is
utilized to gather comprehensive information on user preferences and restaurant
characteristics.

Furthermore, another study explores the use of online reviews, highlighting their
importance in consumer decision-making. Given the overwhelming number of reviews
for various products and services, this study proposes a content-filtering
recommender system that evaluates and scores individual reviews based on restaurant
characteristics and consumer segments. These numeric scores help sort reviews
according to individual consumer preferences, aiding in more personalized and
efficient decision-making. This approach demonstrates the potential of combining
user reviews and machine learning techniques to create robust and user-centric
recommendation systems.

1. Dataset:
Totally it contains 9 csv files including ratings,cuisine,user payment etc. In
that we analyze
mainly rating file which contains 1161 instances and mainly 5 attributes. This
dataset is integrated with
our application so that the machine learning algorithm deploys the suitable results
for the customer.

2. Popularity Based Recommendation:

This is the common baseline approach. This model
not focused on personalized approach but it simply
recommends to the user the most popular thing in
that region or on that occasion or the best thing in
the appropriate restaurants. Hence the consumer
may use the product or may not use the same on
previously. It recommends the current trending or
popular across the region.

3. Collaborative Filtering:
Collaborative filtering is a most recently using
algorithm for recommendation system. This method
uses inputs from the multiple users with similar taste.
It has user-based CF and item-based CF.
This methodology exploits the users underlying
preferences through the analysis of latent features
that define the input values. We used singular value
decomposition collaborative filtering model.

refernce:-
Restaurant Recommendation System Using
Machine Learning Algorithms
Asst. Prof. Sathya Seelan K, Asst. Prof. K. Prabhu
Department of Information Technology,
Sri Ramakrishna Institute of Technology,
Coimbatore,TN,India

research paper no 4

Personalized Food Recommendation System

by using Machine Learning Models
The study of recommender systems reveals their essential role in filtering
information and offering personalized or non-personalized suggestions based on
users' needs. Personalized systems gather user data through surveys or
questionnaires to understand preferences, while non-personalized systems base
recommendations on observed likes or trending content. According to researchers
Anna Gatzioura and Miquel Sanchez, recommendation systems aim to provide valuable
material to users actively engaged with a site. Emerging as a popular research
subject in the mid-1990s, these systems leverage collaborative filtering to enhance
user experience and boost sales for e-commerce platforms by connecting customers
with preferred products. Falling under unsupervised machine learning models,
recommender systems utilize vast amounts of data from comments, reviews, and other
online content. With internet usage surging to 3.2 billion since 1995, the influx
of information can overwhelm users, complicating decision-making processes.
Recommender systems alleviate this challenge by offering tailored suggestions,
enhancing online browsing and aiding informed decisions, as defined by Lior Rokach,
Bracha Shapira, and Francesco Ricci.

algorithm:-

Confusion Matrix
A confusion matrix table is used to assess how well a
machine learning algorithm works.[3] A square matrix
compares a model’s actual and predicted values for a data
set.

Accuracy, precision, recall, and F1 score are just a

few performance measures that can be calculated using the
confusion matrix.

. Scikit Learn
With various supervised and unsupervised learning
methods, and tools for data preprocessing, model selection,
and evaluation, Scikit-learn is a well-known open-source
machine-learning library for Python

. Decision Tree
The decision tree, a non-parametric supervised learning
method, can solve classification and regression problems. It
is arranged in a hierarchical structure with internal roots,
branches, internal nodes, and child nodes.[6] In data mining
and machine learning, the decision tree is a particular type
of predicting model

K-Nearest Neighbor
The assisted learning method K-Nearest Neighbor is
additionally utilized. [8] The K-NN algorithm puts the new
case in the group that most closely resembles the categories
for the existing cases, indicating a relationship between the
new case and the existing [Link] K-NN algorithm saves
every accessible information and categorizes new material
based on similarities.

. Logistic Regression
Logistic regression is a machine learning technique that
fits into the supervised learning group. A predetermined
collection of independent variables is used to predict the
categorical dependent [Link] the outcome for
a binary dependent variable uses logistic regression.
Therefore, the result must be a number that is discrete or
categorical. Instead of the precise values between 0 and 1,

Random Forest
The guided learning approach includes the machine
learning program, Random Forest. It can be applied to ML
issues that combine categorization and regression.[11] Its
foundation is the idea of ensemble learning, a method for
merging various classifiers to handle challenging issues and
enhance model performance.

reference :-
Personalized Food Recommendation System
by using Machine Learning Models
M.S.N.V. Jitendra1, Maddula Lakshmi Jyosna2, Sai Sri Varsha Veeraghanta3, Shanmuk
Srinivas A4, K Bhargav5
1,2,3,4,5 Department of Computer Science and Engineering,GITAM School of
Technology,
GITAM University,Visakhapatnam, AP,India