Hybrid Recommender System for Educational Resources to the Smart

University Campus Domain

Martin Hideki Mensch Maruyama

1 a

, Luan Willig Silveira

1 b

, Jos

e Palazzo M. de Oliveira

2 c

Isabela Gasparini

3 d

and Vin

ıcius Maran

1 e

Laboratory of Ubiquitous, Mobile and Applied Computing (LUMAC), Polytechnic School,

Federal University of Santa Maria, Av. Roraima, 1000, Santa Maria, Brazil

Informatics Institute, Federal University of Rio Grande do Sul, Porto Alegre, Brazil

Universidade do Estado de Santa Catarina (UDESC), Joinville, Brazil

Keywords:

Recommendation Systems, Smart Campus, Collaborative Filtering, Content-Based Filtering.

Abstract:

The development of new cutting-edge technologies in recent years and the ease of access to the internet, the

amount of data circulating on the network have been severely increasing, making it difﬁcult to access quality

information and causing many users to waste their time looking for and ﬁltering through data. Thus, rec-

ommendation systems appears. They are responsible for searching relevant information to the user through

mechanisms capable of recognizing the user’s possible interests and, with the use of recommendation algo-

rithms, bringing the user resources that meet their interests. Actually, recommender systems are applied in

many domains, including news, healthcare, and ﬁnance. Recently, recommender systems have been applied

in smart campus domain, which deﬁnes systems and techonologies to be applied in university campus. From

this scenario, the objective of this study is to develop a hybrid recommender system, attached to a software

architecture, to provide general educational resources to users. The prototype of the architecture was evaluated

using real item data and shown a signiﬁcant accuracy in the recommendation process.

1 INTRODUCTION

With the exponential growth of the media in recent

years, the increase in the amount of data circulating

on internet has become a problem for many areas such

as digital commerce, social networks, entertainment

sites and many platforms in many domains. An ex-

ample of domain which the ammount of data is not

integrated between different systems is in academy or

universities. The access of information that is really

of user interest is frequently difﬁcult in academic en-

vironment. In this context, recommender systems ap-

pear as an alternative to reduce this amount of data

and make the task of searching for a particular item

simpler and faster (Chun-Mei et al., 2021).

Along with this, in recent years, the frequently

changing on the student’s proﬁle has led to a demand

https://orcid.org/0000-0002-2606-581X

https://orcid.org/0000-0002-0187-3554

https://orcid.org/0000-0002-9166-8801

https://orcid.org/0000-0002-8094-9261

https://orcid.org/0000-0003-1916-8893

for new teaching and learning methods to better meet

the student’s needs. Thus, classic classrooms are not

the only spaces learning process (Jord

an et al., 2021)

in the students’ life. Online platforms are not limited

by space or time, in addition, having a multitude of

resources, they have become important and attractive

in recent years (Zhong et al., 2020).

However, as much as these online platforms are

convenient and have improved teaching and learn-

ing management and innovated education in general,

there still issues of getting students to receive per-

sonalized recommendations and making their learn-

ing process more efﬁcient. Because these platforms

have a large amount of information, the time spent by

students in the searching of quality resources is still

a slow and painfully task, because, even though there

is a system that already ﬁlters data, more and more

information are constantly added. On the other hand,

in very new platforms, the small amount of data limits

the students’ learning scope (Meng and Cheng, 2021).

With the development of new technologies such

as IoT (Internet of Things), another area of research

Maruyama, M., Silveira, L., M. de Oliveira, J., Gasparini, I. and Maran, V.

Hybrid Recommender System for Educational Resources to the Smart University Campus Domain.

DOI: 10.5220/0011841900003470

In Proceedings of the 15th International Conference on Computer Supported Education (CSEDU 2023) - Volume 1, pages 47-56

ISBN: 978-989-758-641-5; ISSN: 2184-5026

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

that has become popular is smart campuses, or smart

university campuses. In order to adapt to students’

demands through the use of new teaching technolo-

gies, many universities realized that it was possible

to make changes within their environments using big

data analysis and, consequently, raise the quality of

the services offered, reduce their costs and improve

efﬁciency of the local management of people, data

and general resources (Xu et al., 2018).

Considering the presented context, the main ob-

jective of this study is to model and prototype a soft-

ware architecture that integrate recommenders of dif-

ferent types and make the recommendation of ed-

ucational resources: Courses, mini-courses, video

classes, scientiﬁc articles, lectures, events, theses,

teaching materials, e-books, among others.

The presentation of the paper is structured as fol-

lows: In Section 2 the concepts of recommender sys-

tems and intelligent campuses are presented, it also

present related research focused on these areas. In

Section 3 the proposed architecture and operation of

the recommender systems are presented. The recom-

mendation process is divided into two main ﬁltering

techniques: collaborative ﬁltering (CF) and content-

based ﬁltering (CBF). In Section 4 the results ob-

tained throughout the development of the research are

presented. And, ﬁnally, in Section 5 the conclusions

of this work are presented, which summarizes the in-

formation from the entire research, as well as infor-

mation related to the future directions of the project.

2 BACKGROUND

This section presents the main concepts related to the

proposed software architecture and presents a set of

related work, with the main differences regarding this

proposal.

2.1 Recommender Systems

With the popularization of the internet in the early

90s, the number of active users became extremely

high, while the information circulating through it also

grew exponentially every day, causing an overcrowd-

ing of data in websites, social networks and browsers,

making it difﬁcult for users to access relevant infor-

mation (Meng and Cheng, 2021). Likewise, there was

also great potential on the part of companies to use the

internet as a means of selling products and publiciz-

ing their work, which would attract even more new

users (Cho et al., 2007).

The problem of the massive amount of data on

the internet makes the search for speciﬁc information

an extremely slow and time-consuming task, in ad-

dition, linguistic variability and the use of different

types of languages as well as expressions, slang and

words with more meanings in documents and web-

sites makes it even more difﬁcult for the user to search

for what he/she is really interested in (Benfares et al.,

2017).

An example of this can be the comparison be-

tween social networks and scientiﬁc articles, where

even though the informal language used in a conver-

sation between two users addresses the same topic of

an article, it will not have the information transmit-

ted in the same way or its quality probably it will be

much inferior to the language of the study, impair-

ing its understanding. From this scenario, alternatives

arise to get around the problem of large amounts of

data on the network, which may be the use of artiﬁ-

cial intelligence, automated learning systems or, also,

recommendation systems (Benfares et al., 2017).

The term and concept of recommender systems

was ﬁrst introduced in the 1990s by Jussi Karlgren

(Karlgren, 1990). Recommender systems are respon-

sible for making predictions about the preferences of

a given user over a massive amount of data by cal-

culating the similarity between items, users and the

applicant’s interests. These systems look for afﬁnities

between information in order to identify possible data

to be recommended (Mrhar and Abik, 2019). These

systems have a wide variety of applications and can

appear on sales sites, news, food (Wang et al., 2021),

music (Zhao et al., 2019), social networks, movies

and series (Benfares et al., 2017), services, jobs (Zhou

et al., 2019), tourism (Hu et al., 2017) and also plat-

forms aimed at the academic ﬁeld ((Samin and Azim,

2019), (Uddin et al., 2021)).

It also happens that with the advancement of the

internet and its increasingly wide use in new technolo-

gies, recommendation systems evolve together and

constantly with it due to the fact that new fronts and

study trends emerge with the objective of optimizing

the use of these systems or creating new applications

for them (Dennouni et al., 2018). Initially, for ex-

ample, recommendation systems were focused on ser-

vices and products (e-commerce), later on social net-

works and user contextual information (geographical

location, social relationships, tastes, etc.) and, ﬁnally,

the union with technologies such as IoT (Internet of

Things) and LBS (Location Based Systems) in mo-

bile applications (Dennouni et al., 2018).

Each recommender uses a recommendation tech-

nique to make the best options for items for each

user based on their interactions with other items,

purchases, watched movies, games, proﬁles of other

users, accessed courses, etc. types of food, places

CSEDU 2023 - 15th International Conference on Computer Supported Education

visited, among other parameters. There are several

ﬁltering techniques used to reach a small number of

resources with a very high probability of being rele-

vant to the user, these techniques include algorithms

and programming libraries that together make up the

recommendation system. Among the most famous are

collaborative, content-based and hybrid ﬁltering.

Collaborative ﬁltering takes into account informa-

tion from users with similar interests to ﬁlter items

(Jord

an et al., 2021). For example in a sales web-

site, when a purchase is about to be made, a tab with

the ”other users also bought:” message and then the

related products are displayed, that is, the system rec-

ommends to the user other items that would possibly

be of his interest taking into account what he is buy-

ing and what other users, who have also interacted

with the same product, are interested in.

Content-based ﬁltering aims to recommend re-

sources taking into account the description and pa-

rameters of the items themselves and similar to the

user’s interests (Mrhar and Abik, 2019). In a video

platform, for example, when the user accesses a cer-

tain video, it is also offered to him a series of other

contents with themes similar to the one being repro-

duced. When accessing a video about cake recipes, a

possible item to be recommended would be a pudding

recipe, as both ﬁt into the categories recipe, cooking,

sweets, for example.

Finally, hybrid ﬁltering is the union of two or more

ﬁltering techniques (Jord

an et al., 2021), so that the

advantages and applications of one are able to cover

the disadvantages of the other, causing a variation in

the type of item that is being presented to the user and

not always maintaining the same pattern.

2.2 Smart Campus

The word smart, over the last few years, has accompa-

nied the development of new technologies and is used

in multiple terms of applications such as smart sys-

tems, smartphones, smart homes, smart energy, smart

manufacturing, smart buildings, smart cities, among

others (Zhang et al., 2022).

In addition, there are several deﬁnitions for ”in-

telligent”, which can be the ability to make adap-

tations in response to changing circumstances; abil-

ity to demonstrate intelligence; ability of a system

to convert input data into an action (Imbar et al.,

2020). Smart can also be understood as an acronym

with the following meanings: Self-directed (schools

with cloud-based infrastructure), Motivated (strength-

ening teachers’ skills), Adaptive (encouraging online

classes), Resource (development and use of digital

teaching materials ) and Technology (global reach of

information) (Imbar et al., 2020).

This means that the tools used in education must

be smart for both students and teachers, that their in-

terests in class are shared equally and that teaching re-

sources are made available through technologies that

use cloud services, developing and enriching knowl-

edge. provided by these systems (Imbar et al., 2020).

Smart campus can be described in different ways

depending on the author, but all concepts are valid

and similar. A universal deﬁnition for ”smart cam-

pus” (Chagnon-Lessard et al., 2021) was not found,

however several authors present the following deﬁni-

tions for smart campuses, according to some of them:

• (Du et al., 2016): Smart Campus is the integration

of all kinds of application service systems, creat-

ing a living environment with intelligent learning

and teaching, which is suitable for: management,

teaching, scientiﬁc research and health, and is also

based on IoT;

• (Zhang et al., 2022): Smart Campus is the de-

ployment of advanced information and commu-

nication technologies (ICT) to increase the effec-

tiveness and efﬁciency of campus activities;

• (Xu et al., 2018): Smart Campus is the new direc-

tion of information education. Social networks,

cloud computing, big data, mobile technology,

IoT and other technologies serve as support for

educational informatization, which provide new

ideas for the study of education technologies. The

development of information technologies reﬂects,

mainly, in the development of the understanding

of this information in universities;

• (da N

obrega et al., 2022): Smart Campus

is a higher education institute that creates an

ecosystem through information and communica-

tion technologies (ICT) to achieve sustainabil-

ity using an adaptive and collaborative learning

model to promote a better user experience.

According to (Imbar et al., 2020), a university

can only be called intelligent if it manages to use

its knowledge for study, resolve conﬂicts of interest

between users (students, professors, employees, em-

ployees), and use the intelligence and skills of the

public to contribute to system development. Based

on the deﬁnitions presented, it is possible to say that a

smart campus is a university environment capable of

providing its users with tools, services and resources

that aim to resolve their conﬂicts of interest, through

cutting-edge technologies such as IoT and smart ob-

jects. A list of key characteristics of an intelligent

campus was deﬁney by (Abualnaaj et al., 2020):

Hybrid Recommender System for Educational Resources to the Smart University Campus Domain

• Smart Card or e-Card: Access to classrooms,

laboratories, dormitories, library; digital wallet

payments; data storage control.

• Smart Classrooms: Virtual reality; interactive

and collaborative platforms; remote teaching and

learning; collaborative research.

• Energy Management: Sustainable and smart en-

ergy management systems; use of renewable ener-

gies; smart lighting; electric vehicle charging sys-

tem.

• Adaptative Learning: Personalized teaching

methods; speciﬁc supplementary courses and dis-

ciplines; computerized ﬁtting tests; educational

resource recommender systems.

• Smart Transportation: Smart parking; tracking

of vehicles used on campus; smart navigation.

• Security and Safety: Intelligent security and pro-

tection systems.

• Optimization and Analytics Data Center;

• Smart Facilities Services: Sports centers, li-

braries, restaurants, shops; campus social media.

2.3 Related Work

This section brings an overview of similar works that

ﬁt into the areas of recommender systems and in-

telligent campuses, highlighting the types of recom-

mended items and the ﬁltering techniques used in

each one.

(Ibrahim et al., 2019) proposed a framework

aimed at the academic area that aims to recommend

undergraduate and graduate courses (masters, doc-

torates, among others) for students in general. In

his hybrid system, he chooses to use collaborative

and content-based ﬁltering along with ontology for

extracting and integrating information from multiple

sources.

(Kong et al., 2017) bring in study a model of

collaborator recommendation system, that is, re-

searchers with similar research interests so that other

researchers can be active and help to develop other

scientiﬁc works. The model is based on collabora-

tive ﬁltering, content-based ﬁltering and another type

known as social network-based, creating a hybrid

model.

(Mrhar and Abik, 2019), in turn, proposes a rec-

ommendation system for online course platforms that

can make personalized recommendations based on

each user’s proﬁle. The author makes use of content-

based ﬁltering and deep learning to improve the accu-

racy of the algorithm.

(Xiao et al., 2018) ﬁnally, brings a personalized

recommendation system based on the interests and

history of each user to recommend them didactic ma-

terials and resources necessary for their learning. This

system makes use of content-based and collaborative

ﬁltering.

Table 1: Studies and characteristics of your recommenda-

tion systems.

Studies Recommended

item

Filtering

(Ibrahim et al.,

2019)

Undergraduate

and postgraduate

courses

Ontology

based ﬁltering,

Content-based

ﬁltering, Collab-

orative ﬁltering

(Kong et al.,

2017)

Collaborating re-

searchers

Collaborative ﬁl-

tering, Content-

based ﬁltering,

Social network-

based ﬁltering

(Mrhar and

Abik, 2019)

Online courses Content-based

ﬁltering, Deep

Learning

(Xiao et al.,

2018)

Didactic materi-

als for learning

Collaborative ﬁl-

tering, Content-

based ﬁltering

This research Educational re-

sources (courses,

mini-courses,

video lessons,

teaching mate-

rials, e-books,

lectures, events,

scientiﬁc articles,

theses, similar

user proﬁles,

other educational

platforms)

Collaborative ﬁl-

tering, Content-

based ﬁltering

Based on these and other studies in these areas,

this work aims to develop a platform that would not be

limited to recommending just a few types of items, but

a variety of them, covering both the items discussed

above and many others. Table 1 presents the types of

resources recommended in each study and the tech-

niques used for data processing.

3 HYBRID RECOMMENDATIONS

TO SMART CAMPUS DOMAIN

In this section we present the deﬁnition of the pro-

posed software architecture and the deﬁnition of the

recommendation strategies applied to it.

CSEDU 2023 - 15th International Conference on Computer Supported Education

3.1 The SmartC Software Architecture

The SmartC platform is designed to be a software ar-

chitecture, modeled in a set of services, to provide the

main structures to use different recommender systems

for different types of items. The services and layers of

the architecture were deﬁned specially for the smart

campus domain (presented in Figure 1).

The architecture can be divided into three sections

each responsible for a part of the SmartC system’s op-

eration: the access environment, the recommendation

management environment, and the persistence layer

(da Silva Lopes et al., 2022). In the access environ-

ment, it is the part accessible to users, it is where rec-

ommendations are presented, users can interact with

resources, direct themselves to other university por-

tals, inform their interests, for example, all through

devices such as cell phones. or computer.

In the recommendation management environment

or development environment, only developers have

access to it, because here are all the codes and func-

tions of the system and where new codes are added

or edited and corrected. This environment is also

where the entire process of personalized recommen-

dation for each user takes place, where their informa-

tion is collected and processed and then requests are

sent to the database so that the data obtained can be

worked on by the recommendation algorithms.

Finally, in the resistance layer we have the

database, responsible for storing all system informa-

tion in tables, that is, descriptions of resources, inter-

ests of each user, interacted items, history of recom-

mendations, evaluations, for example, are all saved

in this section and which, when requested, are sent

to the development management environment to start

the ﬁltering and recommendation process.

The recommendation algorithm, in turn, devel-

oped on the platform, aims to recommend to users,

in a personalized way, educational resources based on

topics of interest that the user must inform when ac-

cessing the system. It is a hybrid system because it

has two types of ﬁltering: content-based ﬁltering and

collaborative ﬁltering. Because these two types are

available, it was decided to switch between the ﬁl-

ters every time the user requests new recommenda-

tions, making new resources recommended each time

the system is called and avoiding a certain limitation.

what the user can access. As such, each ﬁlter and its

use is explained in the following sections.

3.2 Content-Based Recommender

In algorithms that use content-based ﬁltering (CBF),

items will be recommended based on the user’s inter-

ests (Thannimalai and Zhang, 2021). This type of ﬁl-

tering also searches for items that are similar to each

other, that is, based on the description of a given re-

source, the system will look for other resources with

similar characteristics to be recommended. In addi-

tion, because this type of ﬁltering depends only on the

user’s interests and the characteristics of the items, it

is already capable of making good recommendations

right from the start without the need for a prior in-

formation base, and it is also very volatile because if

the user’s interests change, the recommendations will

also change (Eliyas and Ranjana, 2022).

For example, on movies and series website, whose

system uses the CBF, recommendations will be made

based on the content that the user has most watched

or watched recently. In terms of security and privacy,

CBF does not require, for example, that the user share

his interests or have to make them public, it is enough

for him to access the contents of the system and his

preference information is already processed and prop-

erly stored (Thannimalai and Zhang, 2021).

In this study, the developed algorithm makes use

of CBF works as follows: initially, the interests in-

formed to the system by the user are stored in ta-

bles that relate the user to each of these topics and,

from this relationship, all the resources that contain

any of these topics are listed; once listed, if they have

not yet gone through this process, any and all text

present in the parameters of each resource is trans-

formed into a string and goes through a textual ﬁlter-

ing technique known as bag-of-words, which removes

from a text all possible keywords (relevant words) and

counts each one, creating a list of words and their

appearances in the text which, in turn, is added as a

resource parameter; after that, this same process is

applied to the list of resources already favorited by

the user, so that a calculation of similarity between

the textual content of each resource with the other is

performed using cosine similarity; ﬁnally, the 25 re-

sources that are most similar to the user’s favorites list

are recommended to the user.

3.3 Collaborative Recommender

Collaborative ﬁltering (CF) is the most successful and

used technology in the area of recommender systems,

its recommendation techniques have a wide range of

applications in the most diverse sectors such as digital

commerce (e-commerce) and social networks (Chen

et al., 2018). The main idea of a recommendation al-

gorithm that uses CF is to recommend items that are

possibly of interest to the user based on their relation-

ships with other users and/or relationships between

(Zheng et al., 2020) items. This type of system is

Hybrid Recommender System for Educational Resources to the Smart University Campus Domain

Figure 1: SmartC software architecture deﬁnition.

based on evaluation matrices where each user informs

how relevant a given resource is for him or her, how

much he or she liked or disliked a certain recommen-

dation and the methods used act directly by process-

ing and computing the information from these ma-

trices to generate new recommendations (Valdiviezo-

Diaz et al., 2019).

However, systems with FC may have impasses

due to problems such as big data, information spar-

sity and cold-start, which severely affect the quality

and accuracy of recommendations (Chen et al., 2018).

Therefore, it is customary to make integrated use of

other technologies and ﬁltering techniques such as,

for example, clustering, Singular Value Decomposi-

tion (SVD), Probability Matrix Factorization (PMF),

recommendation with social trust ensemble (RSTE),

Social rating Matrix Factorization (SocialMF), to

work around these problems (Chen et al., 2018).

CF can be divided into two types of approaches:

model-based and memory-based (Valdiviezo-Diaz

et al., 2019). In a model-based approach, a model

is created from data and then recommendations are

made such that user ratings for certain items are mod-

eled with a set of factors that represent characteristics

of those users and items.

The most popular type of implementation of this

CSEDU 2023 - 15th International Conference on Computer Supported Education

approach is Matrix Factorization (MF), in addition,

this method has achieved better results in terms of

performance and accuracy. On the other hand, in the

memory-based approach, the information to be rec-

ommended is obtained directly from the (Valdiviezo-

Diaz et al., 2019) evaluation matrices. And al-

gorithms that use this approach can be further di-

vided into two types: user-based collaborative ﬁl-

tering (user-based CF) and item-based collaborative

ﬁltering (item-based CF) (Thannimalai and Zhang,

2021).

- User-based CF: makes comparisons between

users with similar preferences based on ratings made

on the same items.

- Item-based CF: Creates a list of items similar to

what the user has previously interacted with or rated.

In this study, the developed algorithm that makes

use of the CF in the following way: in the same way as

the CBF, the user informs the system of his interests;

from that, the system will look for other users who

have the same interests and will create a list for each

one; then, a sum of the topics of both users is made

and, for each common interest, 1 is added to the to-

tal; ﬁnally, the algorithm separates the resources that

would be recommended to the most similar users in a

list, shufﬂes it and returns it to the requesting user.

It is also important to point out that none of these

ﬁlters, both CBF and CF, will fulﬁll its role if the user

does not previously inform the system of his interests.

More speciﬁcally, for CBF, if the user has not favor-

ited any resource so far, the recommendations made

to him will be based only on his interests.

The platform was prototyped using the following

technologies: Angular framework for frontend appli-

cation, Python, Flask, Scikit and Surprise! libraries

for backend appplication and PostgreSQL database.

4 EVALUATION

This section presents the evaluation process of the

prototyped architecture, the list of resources available

in the system and the process of evaluating the accu-

racy of the recommender system based on collabora-

tive and content-based ﬁltering. Figure 2 presents an

example of the UI that shows to user how a recom-

mendation is presented and how the user can interact

with this recommendation. The users can: (i) Visual-

ize the item, (ii) Mark the item to be removed of the

recommended set to the user, (iii) Mark the item as

favorite and (iv) Evaluate the item in a Lickert scale

(from one to ﬁve). All the interactions of the user

with the items are recorded and analyzed by the rec-

ommenders of the software architecture.

4.1 Case-Based Scenario

From focus of the project, a survey of possible

resources to be used in the recommendation sys-

tem was carried out. A search was carried out

for items within the scope of Federal University of

Santa Maria

, on items such as scientiﬁc articles,

professors/supervisors, university courses, disciplines

offered, mini-courses, technical courses and video

lessons. The items were identiﬁed and crawlers were

developed to import the metadata of the items to the

recommender system.

Currently, the dataset is composed of 189 top-

ics of interest and educational resources that cover,

mainly, the study areas of the Cachoeira do Sul cam-

pus (Architecture and Urbanism, Electrical Engineer-

ing, Mechanical Engineering, Agricultural Engineer-

ing and Transport and Logistics Engineering). Table

2 shows the number of topics of interest and resources

divided by category, which today make up the dataset.

Table 2: Educational resources used in the evaluation pro-

cess.

Item Category Number of items

Graduate program work 165

Professor 135

Minicourse 78

MsC. Dissertation 26

Research Project 14

Graduate Dissertation 9

Research paper 4

Extension Project 2

General Information Report 2

PhD. Thesis 1

Media publication 1

Total 437

4.2 Results and Discussion

To assess the accuracy of the recommendation algo-

rithm, ﬁctitious users and educational resources were

created. Being the educational resources (total of 10

resources) related to a topic, as shown in Table 3.

A random number of users (between 20 and 100)

was created for each topic of individual interest and

a random number of users (also between 20 and 100)

related to two topics of interest simultaneously. Table

4 shows the number of users interested in each topic

or group of topics.

It can be seen in Table 4 that groups of users

were purposely created with a common interest in En-

gineering and one of the four topics: Art, Science,

Mathematics and Music.

https://www.ufsm.br

Hybrid Recommender System for Educational Resources to the Smart University Campus Domain

Figure 2: Example of the interface of a recommended item, with the explanation of the possible interactions (in portuguese).

Table 3: Educational Resources.

Id Topic

0 Engineering

1 Math

2 Science

3 Art

4 Music

5 Sports

6 History

7 Geography

8 Literature

9 Philosophy

Through this, it is clear that there is a relation-

ship between users who are interested in engineering

and users who are interested in mathematics, science,

art and music. Therefore, it is expected that the al-

gorithm will be able to identify this relationship and

recommend these educational resources to users with

an interest in engineering only.

Table 5 presents the result of the recommenda-

Table 4: Interest topics.

Interest Topics Number of Users

Art 49

Art, Engineering 77

Science 76

Science, Engineering 55

Engineering 81

Engineering, Math 73

Engineering, Music 94

Sports 90

Philosophy 40

Geography 52

History 66

Literature 38

Math 56

Music 46

tion system based on collaborative ﬁltering, where the

topic column represents the recommended resource in

descending order for a user with an interest in engi-

neering alone.

CSEDU 2023 - 15th International Conference on Computer Supported Education

Table 5: Prediction of the evaluations.

user id topic

0 Engineering

0 Art

0 Music

0 Science

0 Math

0 Sports

0 Philosophy

0 History

0 Literature

0 Geography

It can be seen in Table 5 that the resource with the

best prediction of interest for user 0 was art, followed

by music, science and mathematics, that is, the algo-

rithm proved to be capable of identifying the proba-

ble interests of a user with an interest in engineering

alone.

5 CONCLUSIONS

The introduction of state-of-the-art technologies on

university campuses in order to make intelligent cam-

pus an efﬁcient way to develop and improve the ser-

vices and resources offered by the university, making

its users able to fully achieve their goals. The applica-

tion presented in this study aims to serve as a tool for

the development of an intelligent campus, by recom-

mending educational resources to users based on their

interests, making use of multiple ﬁltering techniques

and libraries in its algorithm. The results of the sys-

tem evaluation demonstrate that the algorithm is able

to make accurate predictions about possible interests

of a user, even if he has informed few or even just one

topic of interest.

As future objectives, it is intended to implement a

resource evaluation system in the algorithm and make

the generated recommendations also take this param-

eter into account. With this, it is expected that the

accuracy of the SmartC system will be even greater

and that the recommended resources will please even

more the tastes of the users.

ACKNOWLEDGEMENTS

This research is supported by CNPq/MCTI/FNDCT

Nº 18/2021 grant n. 405973/2021-7 and CAPES -

Financing Code 001. The research by Jos

e Palazzo

M. de Oliveira is partially supported by CNPq

grant 306695/2022-7 PQ-SR. The research by Is-

abela Gasparini is partially supported by CNPq grant

308395/2020-4 and FAPESC Edital nº027/2020 TO

n°2021TR795. The reasearch by Vin

ıcius Maran is

partially supported by CNPq grant 306356/2020-1

and Fundac¸

ao de Amparo a Pesquisa do Estado do

Rio Grande do Sul (FAPERGS), grant n. 21/2551-

0000693-5 and PIBIC program.

REFERENCES

Abualnaaj, K., Ahmed, V., and Saboor, S. (2020). A strate-

gic framework for smart campus.

Benfares, C., Idrissi, Y. E. B. E., and Abouabdellah, A.

(2017). Recommendation semantic of services in

smart city. volume Part F129474. Association for

Computing Machinery.

Chagnon-Lessard, N., Gosselin, L., Barnabe, S., Bello-

Ochende, T., Fendt, S., Goers, S., Silva, L. C. P. D.,

Schweiger, B., Simmons, R., Vandersickel, A., and

Zhang, P. (2021). Smart campuses: Extensive review

of the last decade of research and current challenges.

IEEE Access, 9:124200–124234.

Chen, R., Hua, Q., Chang, Y. S., Wang, B., Zhang, L., and

Kong, X. (2018). A survey of collaborative ﬁltering-

based recommender systems: from traditional meth-

ods to hybrid methods based on social networks. IEEE

Access, 6:64301–64320.

Cho, D. Y., Kwon, H. J., and Lee, H. Y. (2007). Analysis of

trust in internet and mobile commerce adoption.

Chun-Mei, L., Yi-Han, M., Wei, P., Yan, Q., Jie-Teng, J.,

and Shuo, D. (2021). Personalized recommendation

algorithm for books and its implementation. volume

1738. IOP Publishing Ltd.

da N

obrega, P. I. S., Chim-Miki, A. F., and Castillo-Palacio,

M. (2022). A smart campus framework: Challenges

and opportunities for education based on the sustain-

able development goals. Sustainability (Switzerland),

14.

da Silva Lopes, A., do Nascimento, A. V. L.,

da Rosa Floripes, C. R., Maruyama, M. H. M.,

Lunardi, G. M., and Maran, V. (2022). Smartufsm:

Uma arquitetura de software para suporte a

recomendac¸recomendac¸˜recomendac¸

oes em campus

universit

arios inteligentes.

Dennouni, N., Peter, Y., Lancieri, L., and Slama, Z. (2018).

Towards an incremental recommendation of pois for

mobile tourists without proﬁles. International Journal

of Intelligent Systems and Applications, 10:42–52.

Du, S., Meng, F., and Gao, B. (2016). Research on the

application system of smart campus in the context of

smart city; research on the application system of smart

campus in the context of smart city.

Eliyas, S. and Ranjana, P. (2022). Recommendation sys-

tems: Content-based ﬁltering vs collaborative ﬁlter-

ing. pages 1360–1365. Institute of Electrical and Elec-

tronics Engineers Inc.

Hu, G., Shao, J., Shen, F., Huang, Z., and Shen, H. T.

(2017). Unifying multi-source social media data for

Hybrid Recommender System for Educational Resources to the Smart University Campus Domain

personalized travel route planning. pages 893–896.

Association for Computing Machinery, Inc.

Ibrahim, M. E., Yang, Y., Ndzi, D. L., Yang, G., and

Al-Maliki, M. (2019). Ontology-based personalized

course recommendation framework. IEEE Access,

7:5180–5199.

Imbar, R. V., Supangkat, S. H., and Langi, A. Z. (2020).

Smart campus model: A literature review. Institute of

Electrical and Electronics Engineers Inc.

Jord

an, J., Valero, S., Turr

o, C., and Botti, V. (2021). Using

a hybrid recommending system for learning videos in

ﬂipped classrooms and moocs. Electronics (Switzer-

land), 10.

Karlgren, J. (1990). An algebra for recommendations an

algebra for recommendations using reader data as a

basis for measuring document proximity.

Kong, X., Jiang, H., Bekele, T. M., Wang, W., and Xu,

Z. (2017). Random walk-based beneﬁcial collabo-

rators recommendation exploiting dynamic research

interests and academic inﬂuence. pages 1371–1377.

International World Wide Web Conferences Steering

Committee.

Meng, H. and Cheng, Y. (2021). Research on key technolo-

gies of intelligent recommendation based online edu-

cation platform in big data environment. pages 638–

645. Association for Computing Machinery.

Mrhar, K. and Abik, M. (2019). Toward a deep recom-

mender system for moocs platforms. pages 173–177.

Association for Computing Machinery.

Samin, H. and Azim, T. (2019). Knowledge based recom-

mender system for academia using machine learning:

A case study on higher education landscape of pak-

istan. IEEE Access, 7:67081–67093.

Thannimalai, V. and Zhang, L. (2021). A content based and

collaborative ﬁltering recommender system. volume

2021-December. IEEE Computer Society.

Uddin, I., Imran, A. S., Muhammad, K., Fayyaz, N.,

and Sajjad, M. (2021). A systematic mapping re-

view on mooc recommender systems. IEEE Access,

9:118379–118405.

Valdiviezo-Diaz, P., Ortega, F., Cobos, E., and Lara-

Cabrera, R. (2019). A collaborative ﬁltering ap-

proach based on na

ıve bayes classiﬁer. IEEE Access,

7:108581–108592.

Wang, W., Duan, L. Y., Jiang, H., Jing, P., Song, X.,

and Nie, L. (2021). Market2dish: Health-aware food

recommendation. ACM Transactions on Multimedia

Computing, Communications and Applications, 17.

Xiao, J., Wang, M., Jiang, B., and Li, J. (2018). A person-

alized recommendation system with combinational al-

gorithm for online learning. Journal of Ambient Intel-

ligence and Humanized Computing, 9:667–677.

Xu, X., Wang, Y., and Yu, S. (2018). Teaching performance

evaluation in smart campus. IEEE Access, 6:77754–

77766.

Zhang, Y., Yip, C., Lu, E., and Dong, Z. Y. (2022). A

systematic review on technologies and applications in

smart campus: A human-centered case study. IEEE

Access, 10:16134–16149.

Zhao, G., Fu, H., Song, R., Sakai, T., Chen, Z., Xie, X., and

Qian, X. (2019). Personalized reason generation for

explainable song recommendation. ACM Transactions

on Intelligent Systems and Technology, 10.

Zheng, K., Yang, X., Wang, Y., Wu, Y., and Zheng, X.

(2020). Collaborative ﬁltering recommendation al-

gorithm based on variational inference. International

Journal of Crowd Science, 4:31–44.

Zhong, L., Wei, Y., Yao, H., Deng, W., Wang, Z., and Tong,

M. (2020). Review of deep learning-based personal-

ized learning recommendation. pages 145–149. Asso-

ciation for Computing Machinery.

Zhou, Q., Liao, F., Chen, C., and Ge, L. (2019). Job rec-

ommendation algorithm for graduates based on per-

sonalized preference. CCF Transactions on Pervasive

Computing and Interaction, 1:260–274.

CSEDU 2023 - 15th International Conference on Computer Supported Education