Optimizing Dynamic Multi-Agent Performance in E-Learning

Environment

D. K. Aarthi and E. J. Thomson Fredrik

Karpagam Academy of Higher Education, India

Keywords: Web Usage Analysis, Web Content Analysis, Web Customization, Lingo, E-Learning Platform, HITS.

Abstract: The instructor-centric paradigm has been displaced as the most cutting-edge method of learning with the

introduction of web-based learning and content management systems. For e-learning systems, web mining is

extremely essential. The user can alter the learning setting in a personalized E-Learning system according to

their preferences. A link that receives the most hits will be displayed first in a general search procedure. To

construct a customizable system, user logs must be used to store each user's historical information. The

proposed approach provides a novel viewpoint by combining web usage mining, the HIT algorithm, and web

content mining. It combines user logs and web page hit statistics and contains data that has been clustered

using the Lingo clustering method. We will discuss a method in this article that makes use of content mining

and web usage to personalized e-Learning services. The usefulness and advantages of web mining for e-

learning are examined in this essay.

1 INTRODUCTION

The World Wide Web has evolved into a powerful

and sophisticated medium for the exchange of data.

Different clients who are geologically located in

superior locations must efficiently access the

different data types. The online get-to-log record,

which is a massive store created by the travels of users

with online destinations, can be searched to discover

the client's navigational patterns. Web Usage Data

Mining is a term used to describe the analysis of Web

traffic trace records. The Internet data explosion has

established degrees of search engine popularity.

People are far from content with how the

implementation of Question noting frameworks has

appeared online to help users find more precise

answers to queries made in a previous period of

search engines. This new generation of frameworks

attempts to organize documents, in contrast to

traditional search engines that only use watchwords

to do so. This new generation of frameworks tries to

understand the client's query and suggests some

comparable queries that other people have frequently

raised and for which the framework has the correct

responses. The truth is that most of the time, suitable

responses have been organized or double-checked by

human editors.

This ensures that, if one of the suggested

questions is truly similar to the client's, the appropri-

Figure 1: Structure of Web Scrapping (Mining)

web content

analysis.

ate answers provided by the framework will be

applicable. The underlying premise of such a

framework is that numerous people are frequently

preoccupied with comparable questions. Some web

search engines have developed strategies to suggest

alternative queries to users to solve the problems.

These techniques are used to give clients the option

to include optional related queries in their search

process, either to address their data requirements or to

rephrase their query strategy to find more relevant

search results. The techniques employed in these

Aarthi, D. and Fredrik, E.

Optimizing Dynamic Multi-Agent Performance in E-Learning Environment.

DOI: 10.5220/0012613000003739

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Artiﬁcial Intelligence for Internet of Things: Accelerating Innovation in Industry and Consumer Electronics (AI4IoT 2023), pages 259-264

ISBN: 978-989-758-661-3

259

constrictive business frameworks are often secure,

but be careful that some of the suggested searches

provided by these search engines contain identical

phrases. This may indicate that those proposed

queries will likely be created by straightforward

query expansion techniques. For example, if a user

searches for Yahoo! search engine, the following

related searches are displayed: messenger, best yahoo

mail. Yet, as we can assume, there is a tonne of

additional queries that concern mail but likely don't

directly mention "yahoo" in their term vectors.

2 RELATED WORK

A Progressive Molecule Multitude Enhancer

and Its Versatile Variation by S. Janson

and M. Middendorf

This research presents a different leveled version of

the molecular swarm enhancement (PSO)

metaheuristic. The particles are arranged in a

dynamic chain of importance in the new technique

known as H-PSO, which is used to describe an area

structure. The nature of the particles' currently best-

understood arrangement determines whether they go

up or down the chain of command. This increases the

impact of powerful particles on the swarm as they

advance in the progressive system. In this form of the

H-PSO, the progressive system is perfectly adjusted

as the computation is being done. Another approach

is to assign different behavior to the individual

particles depending on where they fall in the

significance chain. We test H-PSO and its variants on

a commonly used combination of streamlining

capabilities and are compared to PSO's use of several

standard neighborhood layouts.

B Finding and following Several Intuitive

Optima by a Substantial Swarm Model

using Speciation by D. Parrott and X. Li

To deal with the challenges of continuous

improvement and to track many optimums in a

dynamic environment, this work offers an improved

molecular swarm analyzer that applies the species

idea to identify its local best qualities. According to

their similarity, the swarm populace is divided into

species subpopulations in the proposed species-based

molecular swarm streamlining (SPSO). Every species

is worked around an expert particle known as the

species seed. For each of these unique species

bunches

independently, species seeds are chosen

from the complete population at each cycle stage. At

Figure 2: The architecture of the proposed approach.

Figure 3: The mechanism for the browser's search interface.

each stage of the evolution, species are adaptively

framed in accordance with the information gleaned

from the multimodal wellness scene. Species can

continue to evolve towards different optimum states

over increasing emphasis, paying little attention to

whether they are close or global. Our studies using a

powerful SPSO (DSPSO) to track numerous altering

optimal circumstances in a dynamic environment and

the SPSO to locate many ideal conditions in a static

area have demonstrated the SPSO's outstanding

AI4IoT 2023 - First International Conference on Artiﬁcial Intelligence for Internet of things (AI4IOT): Accelerating Innovation in Industry

and Consumer Electronics

260

capability to handle multimodal improvement tasks in

the two scenarios.

C Developing a Customized E-Learning

Framework in Light of Hereditary

Calculation and Case-Study Thinking

Approach by M.- J. Huang, H.- S. Huang,

and M.- Y. Chen

Students can access instructional resources on a

special platform provided by the World Wide Web.

When instructional content is delivered in

hypermedia form in a framework for learning that is

based on the Web, learning transforms into an

activity-driven process. It encourages pupils to look

into independent navigational routes within the

region while studying from numerous resources

throughout the globe. To facilitate Online education

through the Internet and offer flexible teaching

techniques, A number of analysts have been

concentrating on creating e-learning frameworks with

movable learning elements. While providing

customized educational program sequencing

administrations, the majority of customized

frameworks do take student preferences, premiums,

and reading habits into account. Nevertheless, these

frameworks typically fail to consider whether student

ability and the degree of difficulty of the suggested

training materials are compatible. In this manner, our

suggested approach is based on the evolution method

using adaptable electronic testing (CAT). At that

moment, a perfect learning method is created for each

learner using case-based reasoning (CBR) and

hereditary calculation (GA). Three fundamental

promises are made in this essay: (3) The proposed

methodology can develop the right course materials

for students, taking into account individual student

needs, to help them learn more effectively, as

demonstrated by the observational study. This is

accomplished by (1) laying out the case-based

reasoning needed to construct an integrated analysis

or evaluation investigation, (2) outlining the

hereditary-based educational programs sequencing

approach, and (3) outlining the case-based thinking to

build up a customized educational module

sequencing.

D H. Izakian, A. Abraham, and V. Snael's

paper, "Fuzzy Clustering Utilizing Hybrid

Fuzzy c-means and Fuzzy Particle Swarm

Optimization"

An important problem called fluffy bunching is the

focus of dynamic study in a few real applications. The

most effective, straightforward, and straightforward

to use fluffy grouping method is the fluffy c-implies

(FCM) computation. Nonetheless, FCM is easily

caught in surrounding optima and is sensitive to the

statement. Many advancement challenges are

resolved using molecular swarm optimization (PSO),

a stochastic global streamlining method. This paper

proposes a fluffy grouping method based on fluffy

PSO (FPSO) and FCM, utilizing the advantages of the

two computations. Trial findings show that our

suggested method is effective and capable of

revealing outcomes that are inspiring.

E Adaptive Molecule Multitude Enhancement

by Z.- H. Zhan, J. Zhang, Y. Li, and H. S.-

H. Chung

A flexible molecule swarm improvement (APSO) is

presented that emphasizes preferred search efficacy

over conventional molecule swarm streamlining

(PSO). Most importantly, It is capable of doing a

faster mixing rate global search across the whole

search universe. There are two main breakthroughs in

the APSO. To determine one of the other four defined

transformational states, a consistent developing state

estimation approach is first applied. Inquiry, abuse,

combination, and jumping out of all ages, by

assessing population circulation and molecular well-

being. To increase the efficiency and speed of the

assembly process, It makes it possible to code the run

control of algorithmic parameters like quickening

coefficients and idleness weight. When the

developmental condition is at that point, an elitist

learning system is implemented state of delegated

combination. The approach will look for the molecule

that can exit a potential neighboring optimum the

fastest overall. The APSO's performance was

extensively evaluated against 12 benchmark single-

functional and multifunctional capabilities. We'll

think about the effects of elitist learning and

parameter tweaking. The results show that APSO

significantly improves the implementation of the PSO

perspective in terms of blending speed, global

optimality, arrangement precision, and calculation

consistency. Only two new elements are introduced

to the PSO viewpoint by APSO; no further plans or

multifaceted use characteristics are offered.

Multicluster, Rejection, and against

Combination in Dynamic Conditions by T.

Blackwell and J. Branke

Some real problems are dynamic, necessitating a

streamlined computation that can consistently monitor

Optimizing Dynamic Multi-Agent Performance in E-Learning Environment

261

Figure 4: Console of the implemented system.

an evolving ideal over time. So, we examine novel

iterations of molecular swarm advancement (PSO)

that are specially designed to perform excellently in

challenging circumstances. The key idea is to organize

the particle population into a group of interacting

clusters. These swarms are connected locally by a

parameter for avoiding conflict and globally by a

different administrator who forbids intermingling.

Also,

each

swarm

maintains

respectable

level

variation through the use of stimulating or quantum

particles. In this study, which also sets criteria for

choosing the included parameters, the multiswarm

computations are assessed on a range of examples of

the multimodal lively moving pinnacles benchmark.

Further comparisons between the results with other

PSO and developmental calculations are made the

new multiswarm streamlining agent fundamentally

outperforms earlier approaches, closer to the

authoring.

3 WEB MINING PERTINENT TO

E-LEARNING

Web Mining Methodologies

Web mining is a significant subfield in data mining.

Finding important information or trends in web data

is known as web mining. As indicated in the figure, it

may be divided into the three following figure.

1. Web Content analysis

2. Web Structure analysis

3. Web Usage analysis

The practice of extracting significant information

from the text of online sites is known as web content

mining. The web page's content comprises text,

photos, audio, and video, as well as strategies like

grouping or associating websites based on the

relevant branches. It also makes data mining easier.

Web structure mining analyses the web, and one of its

applications is to locate better publications, as well as

websites pertinent to a specific subject or area, or to

find web communities. It is also used to determine the

structure of web pages from web pages by detecting

the current scenario.

Web Structure Analysis

The objective of web structure mining is online

analysis, and one of its applications is to find better

papers. Also, it helps in locating important websites

for a particular subject or branch as well as online

forums.

Moreover, it is employed to make the schema of

web pages apparent.

Web Usage Analysis

Data mining techniques known as "web use mining"

are used to uncover intriguing usage patterns from

web data to better understand and accommodate the

needs of web-based applications. Usage information

keeps track of both the identities of internet users and

the habits of their web browsing. The most important

research initiatives in the area of Web use mining and

customization are also covered. Yet the effectiveness

of this strategy for customization is not as great.

AI4IoT 2023 - First International Conference on Artiﬁcial Intelligence for Internet of things (AI4IOT): Accelerating Innovation in Industry

and Consumer Electronics

262

Web Mining-Based E-Learning System

There are three components to the e-learning system.

User, learning platform, and collection of teaching

resources. A storage server for storing various kinds

of educational resources is called an education

resource library. This web-based system's user is the

learner. The learning platform that provides users

with a web-based learning environment is the web

server.

To create a standard dataset with learning objects,

the administration is in charge. Stop words and stems

are removed from this dataset after preprocessing.

First, a new user registers with the learning portal. As

soon as a user comes into the system using their

unique username and password and performs a

subject-specific search, their search logs are kept on

the server. After that, the hit method is used to give

those logs more weight. In the suggested method, the

lingo clustering algorithm is used to mine content

using preprocessed data. Preprocessed data are then

used to create clusters. User logs, the hits method, and

clustering findings are used to generate the final

results.

4 THE HYPERTEXT-INDUCED

TOPIC SEARCH ALGORITHM

The HIT Search algorithm is a method for locating

documents that are pertinent to a given keyword

topic. When you type a question or term into the

Google search engine, which was developed by

Krishna Bharat while he was working at the Compaq

Systems Research Center, The Hypertext Induced

Topic Search algorithm aids in locating pertinent

keywords whose outcomes are more educated

regarding the search term or question.

The method uses a systematic index of expert

documents. These are pages that are focused on a

single subject and contain links to numerous

unrelated sites on that subject. If authors from non-

affiliated organizations create a page, that website is

considered non-affiliated. The relevancy of the

description text for hyperlinks on expert pages

referring to a particular result page is taken into

consideration when ranking the results.

The system's performance is assessed in various

contexts and in contrast to the earlier approach, which

is only based on the use of mining. Based on a user's

browsing history, the program may be used to provide

customized recommendations. We have covered a

wide range of study topics for a customized E-

learning system in this essay. This work proposes a

unique web mining approach that is based on a

synthesis of web usage analysis and web content

analysis (HITS algorithm), displaying superior

performance improvement than the prior method

based on the current limitations.

5 FUTURE SCOPE

In the future, a personalized curriculum will be

established using the learner's time distribution

pattern, and a feedback system will be constructed

using the learner's social trend. We will also deliver

varied training based on the different levels of

learners. Research can be conducted to create

integration strategies for techniques that can precisely

predict students' success in courses and ways that

assist in choosing a subject or set of courses based on

student's interests.

REFERENCES

A. Rosen, e-Learning 2.0: Proven Practices and Emerging

Technologies to Achieve Real Results. New York, NY,

USA: American Management Association, 2009.

E. Pontes, A. Silva, A. Guel_, and S. Kofuji,

Methodologies, Tools and New Developments for E-

Learning. Rijeka, Croatia: InTech, Feb. 2012.

C.-T. Yang and H.-C. Ho, ``shareable e-learning platform

using data grid technology,'' in Proc. IEEE Int. Conf. e-

Technol., e-Commerce e-Service, Mar./Apr. 2005, pp.

592_595.

C.-M. Chen, H.-M. Lee, and Y.-H. Chen, ``Personalized e-

learning system using item response theory,'' Comput.

Educ., vol. 44, no. 3, pp. 237_255, Apr. 2005.

M.-J. Huang, H.-S. Huang, and M.-Y. Chen, ``Constructing

a personalized e-learning system based on genetic

algorithm and case-based reasoning approach,'' Expert

Syst. Appl., vol. 33, no. 3, pp. 551_564, Oct. 2007.

C.-P. Chu, Y.-C. Chang, and C.-C. Tsai, ``PC2PSO:

Personalized e-course composition based on particle

swarm optimization,'' Appl. Intell., vol. 34, no. 1, pp.

141_154, Feb. 2011.

C.-M. Chen, Y.-L. Hsieh, and S.-H. Hsu, ``Mining learner

profile utilizing association rule for Web-based

learning diagnosis,'' Expert Syst. Appl., vol. 33, no. 1,

pp. 6_22, Jul. 2007.

C. Romero and S. Ventura, ``Educational data mining: A

survey from 1995 to 2005,'' Expert Syst. Appl., vol. 33,

pp. 135_146, Jul. 2007.

J. M. M. Vazquez, J. A. O. Ramirez, L. Gonzalez-Abril, and

F. V. Morente, ``Designing adaptive learning itineraries

using features modeling and swarm intelligence,''

Neural Comput. Appl., vol. 20, no. 5, pp. 623_639, Jul.

2011.

L. de Marcos, J.-J. Martínez, and J.-A. Gutierrez, ``Swarm

intelligence in e-learning: Swarm intelligence in e-

Optimizing Dynamic Multi-Agent Performance in E-Learning Environment

263

learning: A learning object sequencing agent based on

competencies,'' in Proc. 10th Annu. Conf. Genetic Evol.

Comput., Atlanta, GA, USA, Jul. 2008, pp. 17_24.

P. Brusilovsky and C. Peylo, ``Adaptive and intelligent

technologies for Web-based education,'' Int. J. Artif.

Intell. Educ., vol. 13, no. 2, pp. 159_172, Apr. 2003.

R. Stathacopoulou, M. Grigoriadou, M. Samarakou, and D.

Mitropoulos, ``Monitoring students' actions and using

teachers' expertise in implementing and evaluating the

neural network-based fuzzy diagnostic model,'' Expert

Syst. Appl., vol. 32, no. 4, pp. 955_975, May 2007.

Y.-M. Huang, J.-N. Chen, T.-C. Huang, Y.-L. Jeng, and Y.-

H. Kuo, ``Standardized course generation process using

dynamic fuzzy Petri nets,'' Expert Syst. Appl., vol. 34,

no. 1, pp. 72_86, Jan. 2008.

[14] A. Carlisle and G. Dozier, ``An off-the-shelf PSO,'' in

Proc. Part. Swarm Optim. Workshop, Apr. 2001, pp.

1_6.

Y.-M. Chen, T.-Y. Chen, H.-C. Chu, and K.-C. Su,

``Ontology-based adaptive dynamic e-learning map

planning method for conceptual knowledge learning,''

Int. J. Web-Based Learn. Teach. Technol., vol. 11, no.

1, pp. 1_20, Jan. 2016.

K. Almohammadi, H. Hagras, D. Alghazzawi, and G.

Aldabbagh, ``A survey of artificial intelligence

techniques employed for adaptive. Educational systems

within e-learning platforms,'' J. Artif. Intell. Soft

Comput. Res., vol. 7, no. 1, pp. 47_64, Dec. 2016.

T. Gopalakrishnan and P. Sengottuvelan, ``A hybrid PSO

with Naïve Bayes classi_er for disengagement

detection in online learning,'' Program, vol. 50, no. 2,

pp. 215_224, Apr. 2016.

S. Janson and M. Middendorf, ``A hierarchical particle

swarm optimizer and its adaptive variant,'' IEEE Trans.

Syst., Man, Cybern. B, Cybern., vol. 35, no. 6, pp.

1272_1282, Dec. 2005.

T. Blackwell and J. Branke, ``Multiswarms, exclusion, and

anti convergence in dynamic environments,'' IEEE

Trans. Evol. Comput., vol. 10, no. 4, pp. 459_472, Aug.

2006.

D. Parrott and X. Li, ``Locating and tracking multiple

dynamic optima by a particle swarm model using

speciation,'' IEEE Trans. Evol. Comput., vol. 10, no. 4,

pp. 440_458, Aug. 2006.

AI4IoT 2023 - First International Conference on Artiﬁcial Intelligence for Internet of things (AI4IOT): Accelerating Innovation in Industry

and Consumer Electronics

264