See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/366521415

Understanding Query Intention in Search Queries of Learners in Blended

Learning Environments

Conference Paper · October 2022

DOI: 10.1109/ICCCMLA56841.2022.9989024

CITATIONS READS
0 80

2 authors:

Natarajan Vivekananthamoorthy Sankar Shanmuganathan

SRM Institute of Science and Technology KCG College of Technology
14 PUBLICATIONS 85 CITATIONS 20 PUBLICATIONS 93 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Natarajan Vivekananthamoorthy on 04 January 2023.

The user has requested enhancement of the downloaded file.

 2022 IEEE 4th International Conference on Cybernetics, Cognition and Machine Learning Applications (ICCCMLA)

Understanding Query Intention in Search Queries of

Learners in Blended Learning Environments
Vivekananthamoorthy Natarajan Sankar Shanmuganathan
Computer Science and Engineering Computer Science and Engineering
Hindustan Institute of Technology and Science KCG College of Technology
Chennai, India Chennai, India
[email protected] [email protected]

Abstract— In spite of the flexibility of learning, at any time, web to arrive at an appropriate ecommerce site. These e-
any place, and any pace, the student engagement decides success learning sites provide educational content mainly to promote
of the e-learning portals today. The `Search` plays vital role in student engagement and to help students to fulfil their learning
directing the learners to land in the correct web portal that can intention of enriching their learning experience which is
satisfy their needs and goals in the blended learning mandated by their academic needs in order to complete their
environments. The objective of this experimental study is to courses successfully.
collect the user behavior of the online learners by using the web
analytic tools `Google Analytics` and the `Google Search An experimental website is maintained by the authors
Console` and the faculty and web administrators can verify the where educational contents for courses are uploaded by the
fulfilment of the learner’s search intent. The datasets pertaining faculty that meets the learning needs of engineering students
to the user behavior are collected from the `Search Console` of affiliated institutions of an university in Tamil Nadu, India.
data in the period Jan-Dec-2020 and analyzed using the metrics
such as position, impressions, clicks. The metrics ‘Page views’
and ‘Total time on site’ are indicators of students’ intention to
use the educational web site to achieve their academic goals like
improving their grades in the final examination. The metrics
such as Page views can predict the exam score using a linear
relation. The multiple regression analysis done on the Google
analytic metrics of the experimental data confirmed that these
metrics can predict the exam score in the blended e-Learning
portals and the results are statistically significant. Individually
the learners` performance has been improved with the increase in
Pageviews reported by the web analytic reports. The click through
data metrics helps the faculty to make improvement in the course
website so as to ensure that the learners’ needs are successfully
met by the online experimental portal.

Keywords— Search queries, Click through rate, Blended

learning, Web analytic tools, Query intention, Learner
engagement, Google Analytics, Google Search Console

I. INTRODUCTION
E-learning platforms serve the diverse educational needs Fig. 1. E-Learning Framework
of students in recent times with the flexibility of learning at
any time, any location according to their convenience. There This website had undergone many improvements in recent
are many universities that conduct prescribed courses with times and still satisfying learners’ educational needs as on
syllabus and regulations and students from many affiliated today [1,2].
institutions enrol themselves in these courses. The institutions
run these courses for the students of their institutions and Figure 1 shows an e-learning Framework which
many instructors teach these courses according to course explains the interactions taking place in e-learning portals.
plans. The final examinations are planned and conducted by The UML diagram depicts the interaction of actors comprising
the universities. Many of the courses are common to students of Student, Faculty, Course Administrator, other Stake holders
belonging to different affiliated institutions. The syllabus, and Technology Providers. Google Analytics, a web analytic
structure and the course code are the common denominator API has been integrated with the online learning portal.
across institutions. In blended learning environments, the Google analytics can be used by e-Learning course
students look for additional learning resources in order to administrators to collect statistics and visual reports about the
facilitate improve their learning performance as well as visitors to their website and the usage behavior of learners [3].
fulfilment their learning goals. The students discover their This study focuses on understanding the web usage pattern of
favourite educational website by searching the web using the online learners. The web usage data by the learners were
keywords such as ‘Course ID’ and ‘Course Name’ etc. This is collected from 01st January, 2020 to 31st December 2020
similar to a consumer’s intention to search for products in the from the reports generated by the web analytic software

978-1-6654-6246-4/22/$31.00 ©2022 IEEE

308
 2022 IEEE 4th International Conference on Cybernetics, Cognition and Machine Learning Applications (ICCCMLA)

‘Google ‘Google Analytics’ and ‘Google Search Console’. intent in web search queries can be classified as Navigational,
Different datasets were generated as ‘csv’ files and further Informational, and Transactional [5]. Researchers made
analysis are done on the data using Pandas Library functions empirical study of E-Com search queries in order to discover
and Python code. The student interaction data of the user behavior patterns and arrived at taxonomy of E-Com
experimental website as reported by the web analytic reports queries which consisted of five clusters [6]. Research studies
is examined to identify patterns in web usage. stipulated that a product search should satisfy competing goals
of a customer and business goals of a product supplier [7].
A. Query, Impressions, Position, and Clicks This is translated as a product search should satisfy both, the
Figure 2 shows a data frame `df` consisting of rows and customer’s needs of what the business offers and the business
columns and it was imported from a dataset from the Search needs of converting customer intent into purchases. Studies
Console Report pertaining to users’ Queries related to seeking on online learning concluded that there exists lack of
information on Course Content. For example, consider a understanding on learner expectations and usage intent of
mobile user, who had made a ‘Google Search’ and typed the learners in accessing material from online content providers
text ‘professional ethics notes’ and submitted the query. Then, [8]. These studies focused on comprehending usage intent and
the user clicks an ‘URL’ from the Search Engine Result Page factors such as ease of use, as well as usefulness of mobile
(SERP) and it lands him in the desirable web page ‘ge6075- devices. The measure for system use can be magnitude of use,
professional-ethics-in-engineering’. Each row contains the frequency, nature of use, appropriateness, extent, and purpose.
terms ‘Search query’, ‘Position’, ‘Impressions’, ‘Clicks’, and
‘Click through rate (CTR)’. Also derived columns were C. Web Analytics
arrived at such as ‘Ngram’, and ‘Cosine Similarity’ from the Web Analytics is the objective tracking, collection,
query string. These columns are useful metrics which can measurement, reporting, and analysis of internet data to
impact Click through rates and provide insights for the web optimize web usage [9]. According to web analytic
designers in order to improve site content. Also the metric researchers, the website owners use the web analytics to better
‘Position’ is the average position of the ‘URL’ appearing in understand their visitors [10]. Many tools allow users to track
the search engine result page which can impact the CTR. data like number of visitors (e.g., total, new, returning,
unique), visitor technology (e.g., monitor type, operating
systems, browsers), locations (e.g., country, state, city),
network hostnames, and search keywords. A case study on
click stream data analysis of an airline website highlight that
web analytics can be used to understand customer behavior to
improve content, to effect design changes and to enhance
performance of website [11]. Many research studies
emphasize that the use of analytics in learning can bring
concrete benefits for students, teachers, and institutions.
Online learners leave foot prints, and web analytics can
capture and gather these data, then analyse and provide
meaningful insights and visualizations for institutions,
teachers, and learners [12]. The use of web analytic software
Fig. 2. Query, Position, Impressions, and Clicks
in school learning system can use regression and classification
techniques to predict student learning outcomes [13]. The
II. LITERATURE REVIEW student data on online learning behavior helps teachers for
A. Search for Learning content improving their instructional practice [14]. The web analytics
The search for learning content by the students has two is the practice of web traffic analysis and it provides marketers
stakeholders namely, the learners and the content providers. for proving return on investment [15].
The learners visit an e-learning site in order access the D. Google Analytics
educational content which is more appropriate for their Google Analytics is a web analytics service offered by
educational needs and can help in achieving their educational Google Inc. that tracks website traffic data and provides key
goals. The goals can be translated into enhancement of statistics and visualizations about website visitors and online
knowledge in the subject matter and the learners can appear usage [16,17]. Google Analytics categorizes the website
for the examination with more confidence. Research studies statistics into three major types: (1) Audience data, (2)
on web searches emphasized the need for identifying the goal Acquisition data, and (3) Behavior data. Audience data
behind the user’s query in order to improve the quality of provides vital information about the users visiting the website,
search engine’s results [4]. The student goals can be their country, city and user types like new or returning etc.
understood by collecting data on their web usage behavior and Also, information on how the users are accessing the website
further analysis of the data and their interpretations. The web like device names, browser names, and operating systems
analytics reports summarize user behavior in terms of search used are also captured. The Behavior Data depicts the user
queries, clicks, page views, and dwell time etc. interactions with the website and are measured using the
B. Query Intent metrics such as ‘Page views’, and ‘Time on page’.
The aim of this study is to understand the query intent of E. Goal Conversion
the learner. It is answering the question why learners are Goals in Google Analytics helps a website owner to keep
visiting e-learning sites. Researchers highlighted that the user track of whether the website is effective and the web content

309
 2022 IEEE 4th International Conference on Cybernetics, Cognition and Machine Learning Applications (ICCCMLA)

design has improved the website performance and its intended III. RESEARCH MODEL
purpose is met. A goal conversion (or goal completion) takes The figure 3 shows the System Architectural Model of user
place when intended actions like landing in a desired course interactions in the e-Learning portal and the framework for
web page of an e-learning site denoted by a specific URL. analyzing the Google Analytic Data using various regression
Research studies highlight that the Key Performance models. The click stream data is uploaded in Jupyter Notebook and
Indicators (KPIs) are used to measure the success of any the statistical analyses were done using python code. The outcome of
website [18]. Some useful KPIs that are used in e-commerce these statistical techniques are various statistical scores such as the
sites are click through rates, Time on Site, Page views per Correlation coefficients, and the Rsquare values etc. Multiple
Session, and Bounce rate etc. regression analysis using the ANOVA method has been done so that
F. Google Search Console the prediction of Exam Score in online portals can be done using the
Google Search Console provides key performance metrics predictors such as Clicks, Page views, and sessions. The faculty
such as clicks, impressions, click through rates, and position launches a course, make announcements and plan
in search results and these data aid the website owners interventions in order to make the course more effective. The
improve performance of their websites. For example, it students browse through the course material and attend quiz
provides information on search queries that can make the users etc. For this study, the search query data and the student
landing in a particular landing page of a website. Google interactions with the experimental e-learning portal pertaining
Search Console offers more suggestions, highlighting to a period of 1st Jan, 2020 to 31st Dec 2020 were used. The
problem areas and opportunities for improvement. The learners discover the educational resources from this web site
combination of Google Search Console, Google Website based on their learning needs. The intent for search for a
Optimizer, and Google Analytics is the efficient way to start particular course arises mainly due to the factors such as the
improving the website [19]. start of the Course, assignment deadlines and approaching of
end semester exams. The schedule depicts the start date, the
G. Web Usage Pattern Analytics end date and the exam date for the courses and it need to be
According to [20], the data collected from the student adhered by the institutions, instructors and the students. The
behavior patterns in online learning portals, increased activity course interaction of students with the experimental web site
was observed before deadlines, in the days leading to up to an happens within the period during which a course is active.
exam. These activities highlight that the students are serious in This study has analyzed the click stream data of student
exam preparation by reviewing the lecture notes as the exam interaction with the course web site and the results could be
dates are coming nearer. The metrics such as the number of presented as reports, charts, graphs, and tables. ‘Google
‘study sessions’, and the ‘number of clicks’ are strong
predictors for a good score in the ensuing exam. Time tracking
of learners’ behavior helps in assessing the learners’
performance in the learning environments [21].
H. Learner Needs and Search Queries
Search plays an important role in fulfilling users’
information needs [8]. Software Developers often use a search
engine for various information needs, including finding
reusable code snippets, understanding APIs, locating
reference examples, learning unfamiliar concepts,
remembering syntactic details, identifying appropriate third-
party libraries, and debugging [22]. In the blended learning
environments, the students look for additional learning
resources in order to facilitate fulfilment of their learning
goals. There are many e-learning sites that provide
educational content, and the students can access the
Fig. 3. System Architectural Model
educational content at any time. The students discover their
favourite educational website using searching using course
code and course title. This is similar to product search in an e- Analytics’ and ‘Google Search Console’ which have been
commerce site. The aim of these e-learning sites is mainly to integrated with the experimental web site and the click stream
promote student engagement and satisfying their learning data has been automatically collected in different granularities
needs and goals. In order to understand the search intent of the were loaded in the Python Lab Notebooks for further analysis.
users, 150,000 search queries and associated result clicks were A. Research Questions
collected and analyzed in an experimental study [23]. RQ1: Does the search queries of learners in online
The search engines often use metadata such as previous learning make the target URL to appear in the first few lines
search history, URL clicks to understand users’ search intent. of search engine result page?
Researchers highlight the software developers look for online RQ2: Does the Multiple Regression Model with the
resources and Google search engine provides support in their independent variables Clicks, Pageviews, Sessions, and Pass
search task resulting in productivity improvement in the count and the dependent variable Total Quiz Score provide a
process [24]. good fit for the data?

310
 2022 IEEE 4th International Conference on Cybernetics, Cognition and Machine Learning Applications (ICCCMLA)

B. Hypotheses
This experimental study focuses on student engagement
and enhancement in learning outcome of online students in e-
Learning portals in blended learning environments. Students
acquire knowledge in online learning portals by way of
searching for the learning resources in the web. The first
hypothesis has been framed in such a way as to understand the
search behavior of young learners pursuing education in higher
educational institutions as well as analyzing the how a student
query lands him in a desirable web site. The google search
console data captures metrics such as Query, Position, Clicks,
and landing pages. The learner lands in a desired web page Fig. 6. Age group wise Sessions Fig. 7. Gender wise
Sessions
more often, if the position of the URL happens to be situated
in the first few positions of the search engine results page. If
the cosine similarity between the search string and the URL is B. Search Query of Users and Results
high, the user clicks could be more successful. The second The objective of creating educational content for a course
hypothesis has been framed in order to prove that the web and uploading in the experimental website by the instructors, is
analytic metrics such as Clicks, Page Views, and the No of for promoting easy access to students to view the course
Sessions can predict the learning outcome of online students material at any time and at any place and at any pace. The new
with the target variable as Exam Score. user to the course web site usually makes queries and arrives
H1: H0 - The URL associated with the first few positions of at the specific course web page, by clicking the desired URL
search engine results will not direct the online learners to land of choice listed in Search Engine Result Pages (SERP). The
in their desired landing pages. arriving at a correct web page of choice is termed as goal
H2: H0 - The prediction of Y-Value, Total quiz score of conversion. The web administrators can get the statistics such
online learners using the predictor variables Clicks, as the ‘number of page views’ of users reported by the web
Pageviews, Sessions, and Pass count will not result in a good analytic software and it is deemed as one of the ‘goal
fit for the Multiple Regression Model used in this study. conversion’ in online learning portals.
Figure 8 shows month wise conversion (Page Views)
IV. RESULTS generated for five courses in the year 2020. It can be seen that
V. Demographics of Users there is hike in Page views for the course ‘Knowledge
Figure 4 shows the age group wise user summary of Management’ (denoted by course-id: ‘it6011’). The peaks can
learners who have accessed the experimental web site for be attributed to the fact that large number of students intensely
learning. It can be noticed from the chart that the maximum browse the course material as the exam deadlines are
users who were using the course web site were in the range approaching. The user interaction data with the experimental
18-24 years. Figure 5 depicts gender wise segregation of users web site for the period starting with 1st Jan-2020 and ending
according to the devices they used for online learning which with 31st Dec-2020 has been captured by Google Search
are either desktop or a mobile device. It can be seen that the Console. The data is available in different granularities and
younger learners and male learners use mobile devices more
rather than desktops.

Fig. 4. Age group wise Users Fig. 5. Gender wise Users

Figure 6 depicts the number of sessions used by the

learners for the three courses namely the System Software,
Professional Ethics (GE6075) and Knowledge Management
(IT6011). The major participating students in the course
sessions are in the age group 18-24. From figure 7, it can be
Fig. 8. Month wise Goal completions - 2020
noticed that large number of male students has accessed the
course material for the above courses of the experimental web
site.

311
 2022 IEEE 4th International Conference on Cybernetics, Cognition and Machine Learning Applications (ICCCMLA)

contains metrics such as Impressions, Clicks, Click through the number of pages accessed is contributed by increase in
Rates, and Position. number of sessions and this happens during the day prior to
university exam for that course.
The metrics ‘Page views’ and ‘Total time on site’ are
indicators of students’ intention to use the educational web site
to achieve their academic goals like improving their grades in
the final examination. It can be inferred these charts that the
‘usage intention’ and the actual ‘site use’ is high for certain
course like ‘it6011’ as it is evident from high peaks in ‘page
views’ noticed in the month of Sep -2020. The Google analytic
report confirms this as can be seen in figure 11 that the high
peak occurred on a single day (Sep, 27th 2020 – Pgviews 348).

Fig. 9. Query-Position-Clicks

Figure 9 shows plot of Clicks and Position for each Query

for the course ‘Knowledge Management’ (IT6011). It can be
seen that certain queries create impressions in the first few
positions on the SERP, get a large number of clicks. The scores
for predicting CTR (With train-data, test-data, and test-size =
0.5) using Linear regression are 0.1877 and 0.1710, whereas
the random forest regression gives a score 0.9494 and 0.6778.
The multiple regression results indicated that the model Fig. 12. Quiz score of respondents for the course KM (2017 – 2020)
explain 56.2% variance and the model was a significant
predictor of CTR, F(2,31) = 19.92, p = 0.000. The final Figure 12 depicts frequency plot of quiz score of the
predictive model was CTR = 17.38 – 0.079* Impressions + respondents who participated in the quiz test in the period
2017-2020. Figure 13 shows month wise page views,
aggregate quiz score, and average time on site plotted for the
year 2020. It can be inferred that the page views and quiz score
are correlated and it can be apparent on viewing the figure 13.

Fig. 10. Click through rates for a) Position, and b) Cosine similarity

0.4545* Clicks.
Figure 11 shows the web usage pattern of student for the
period covering the year range 2017-2020. Fig. 13. Month wise Page Views and Quiz Score
(Jan 2020 – Dec 2020)
C. Web Usage Pattern of Users
It can be seen that patterns can be observed on students’
web access behavior showing peaks and valleys on the page
views as reported by “Google Analytics” as the students
browse the course material for the course ‘Knowledge
Management’(Course Id = ‘IT6011’). It can be inferred that

Fig. 11. Time wise pattern of Page views (Year 2017 – 2020) Fig. 14. Day wise Clicks, Pageviews vs Quiz Score for the month

312
 2022 IEEE 4th International Conference on Cybernetics, Cognition and Machine Learning Applications (ICCCMLA)

Figure 14 gives the pattern of the click count, the total the URL in the search engine results pages (SERP). The
pages viewed, and the sum of quiz score of the participants each regression plot highlights that larger number of clicks happen
day in Sep-2020. at a lower position in the SERP.
It can be seen that the peaks of web analytic metrics can TABLE II. QUIZ-SCORE ANOVA
be correlated to the number of quiz respondents, the number Significance
df SS MS F
of pass percentages. It can be inferred that the performance F
improvement in learning happens and it can be attributed to Regression 4 110411.7 27602 399.5 2.96E-14
the high percentages of course material pages viewed by the Residual 14 967.2 69.08
Total 18 111378.9
learners. Also, the time spent on the course resource pages also
high indicating the students are serious in their online learning
behavior. Based on a regression analysis, the regression line has the
A. Multiple Regression Analysis equation: Clicks_Tot = - 34.5 x Position + 572.4. The

The dataset pertaining to the web usage pattern of students TABLE III. QUIZ-SCORE REGRESSION COEFFICIENTS
contains features such as Pageviews, Clicks, Dwell-time, and
sessions etc. and a multivariate model is constructed using the Coefficients
Standard
t Stat P-value
feature-set mentioned to predict the learning performance of Error
Intercept 3.9427 2.4230 1.6271 0.1259
the students with the assumption that the quiz-score is a
Pass_count 16.1161 1.6292 9.8918 1.07E-07
measure of students’ knowledge gain obtained after querying Pgviews 0.5361 0.2638 2.0323 0.0615
and accessing the experimental website. Figure 15. shows a Clicks 0.0093 0.0880 0.1065 0.9166
scatter plot between Page views and Total quiz score. A Sessions -0.7493 0.3800 -1.9716 0.0687
regression line is fitted for the data and now Page views can
predict the quiz score using a linear relation.
correlation coefficient is -0.917, which signifies a strong
negative correlation. The coefficient of determination is
0.841, indicating that 84.1% of the variation in the data is
determined by the regression line. The p-value = 0.0 which is
less than 0.05 and it provides enough evidence to reject the
null hypothesis H0 : H1.

Fig. 15 How Page views influence Quiz Score of respondents

A multiple regression analysis is done on the collected data

and the results are shown in table I, table II, and table III.

TABLE I QUIZ-SCORE - REGRESSION

STATISTICS
Multiple R 0.995648404
R Square 0.991315745
Adjusted R Square 0.988834529
Standard Error 8.311967377 Fig. 16. Position vs Clicks
Observations 19
V. DISCUSSIONS
The regression model arrives at a regression equation A. Demographic Users
Total_Quiz_Score = [3.942 + 16.116 x Pass count + 0.536 x
Pgviews + 0.0093 x Clicks – 0.749 x Sessions] which is a good Figure 4 to 7 shows the demographics of the users of
fit for the experimental data. In table I, The R-squared value the experimental e-learning portal and the users are
of 0.9913 indicates that the proposed model accounts for about segmented into the following categories: (1) Age group (2)
99.13% of the dependent variable’s variance. R Square is a Gender (3) Device categories and (4) Course selection. The
statistic indicates the percentage of the variance in the students in the age group 18-24 are the major users of the
dependent variable(quiz-score) that the independent variables experimental e-learning portal and the students prefer
(Pass count, Pgviews, Clicks, and Sessions) explain mobile devices more for accessing the web pages rather
collectively. Since the p-value = 2.96E-14 < 0.05 as shown than the desktop. Also, majority of male students prefer
in table II ( ANOVA output) , the Null Hypothesis (H2 : H0) is mobile devices to access the learning portal. Also, it is
rejected for RQ2. The model explains 99.56 % of variance and apparent that majority of users in the age group 18-24 prefer
was a significant predictor of Quiz-Score. Figure 16 shows the to access only certain courses frequently compared to other
total clicks done by the users is dependent on the position of courses. Due to Covid-19 Pandemic spread, nationwide

313
 2022 IEEE 4th International Conference on Cybernetics, Cognition and Machine Learning Applications (ICCCMLA)

lockdown announced by Union Government resulting in F. Customer Search Behavior

schools and college being shut. Most of the educational A research study made on online customer search
institutions resorted to online learning and majority of the behavior on a large e-commerce website adopted a modern
students preferred to use smart phones for online learning machine-learning technique, random forest, as well as
[26,27]. Based on analyzing the survey responses, the logistic regression to predict e-commerce purchases [29].
demographic data revealed that the mean age of the Their results show that the random forest model performs
respondents as 23. Many research studies done on mobile better with a very high accuracy rate (76%) in predicting
learning adoption highlight that mobile learning could customers who will purchase the item they browsed.
increase productivity by providing just-in-time knowledge Among all the predictors, page and session dwell time, and
to the learners [28,29]. click position are the strongest influential factors for the
B. E-learning Initiatives of Government of India conversion behavior.
Government of India’s major e-learning initiative in VI. CONCLUSION AND FUTURE WORK
recent times is ‘SWAYAM’ (Study Webs of Active
Learning conducted in for Young Aspiring Minds) and it is Patterns are seen on the web usage behavior of the online
an integrated platform for online courses in the education blended learning educational portals. The web analytic
sector of India [30,31,32]. Through exams conducted in software “Google Analytics” reported the web access
2018-2019 on SWAYAM 16% and 21% of India’s public behavior showing peaks and valleys on the page views
higher educational institutes have earned credits [33]. pertaining to the period covering the year range 2017-2020. It
was inferred that the number of course pages accessed is
C. Web Usage Pattern contributed by increase in the number of sessions and this
According to [20], the data collected from the student happens during the day prior to the final qualifying exam for
behavior patterns in online learning portals, increased that course. The metrics ‘Page views’ and ‘Time on site’ are
activity was observed before deadlines, in the days leading indicators of students’ intention to use the educational portal
to up to an exam. These activities highlight that the students in order to achieve their academic goals like improving the
are serious in exam preparation by reviewing the lecture grades in the final qualifying exam. The performance
notes as the exam dates are coming nearer. The metrics improvement in online learning happens and it can be
such as the number of ‘study sessions’, ‘the number of attributed to the high percentage of course pages viewed by
clicks’ are strong predictors for a good score in the ensuing the learners. Cosine similarity between Search query terms
exam. In a study on time tracking of students of online
and URL provide insight for improving web content for
learning, the web analytic tool Google Analytics was used
and it provided metrics such as time spent by the learners in successful conversion rates. Multiple Regression analysis
different granularities [21]. The data collected by the done on Google Search console data pertaining to the period
‘Google Analytics’ on the web usage behavior of users for 01-Jan-2020 to 31-Dec-2020 which was collected from the
the period 1st Jan, 2017 to 31st Dec, 2020 on the experimental website and the model was a significant predictor
experimental website is shown as a graph in Figure 11. It of Click through rate(CTR) and the model explain 56.2% of
can be noticed that the graph contains peaks and valleys and variance of target variable CTR (F(2,31) = 19.92, p = 0.000).
the spike in activities can be noticed on Sep 27th, 2020 with Random Forest Regression reported R-Square score of 0.9913
348 pageviews. and an accuracy score of 76.3% respectively. This research
work reinforces that the student online learning performance
D. Search Query and Intention can be predicted based web analytic metrics such as
Understanding user intention in Search Queries can help ‘Pageviews’ and ‘Clicks’. The students’ learning performance
improve web sites in order to satisfy the user needs and goals. improvement was taking place as indicated by enhancement
Automatically discovering user’s intentions is a key challenge in quiz-score as confirmed by the multiple regression analysis
for search engines as they can improve user’s experience by done. The predictor variables are ‘Pass-count’, ‘Page views’,
producing results that are beneficial to the users satisfying ‘Clicks’. The model was a significant predictor of Quiz-Score,
their specific needs [34]. The search engine task involves
F(14,4) = 399.53, p = 0.000. The present study used only a
critical analysis of user intentions so that the user’s clicks
limited web analytic dataset pertaining student interactions related
lead them to the target website. Physicians express their
intentions by querying on text-based content in Electronic to few courses that was deployed in the experimental website.
Health Records and it helps communication across medical The future work involves improvement in the experimental
professionals’ [37]. study with involvement of large number of website users and
E. Search Engine Results Page the e-Learning framework to be validated for repetition in
learning performance enhancement as stipulated by the
According to [35], the extended snippets introduced by Research Model.
Google in 2012 were to show the most important
information like a Knowledge Graph on the Search Engine REFERENCES
Results Page (SERP). It is desirable to get a precise and [1] N. Vivekananthamoorthy, “Enhancing effectiveness of e-learning
concise answer to the user’s query which fulfils the user’s frameworks: Determinant Factors and Driving Forces,” PhD Thesis,
information needs. Putting the right keywords in the page Hindustan University, Chennai, 2017.
URL will help it to be considered as a source for direct [2] N. Vivekananthamoorthy, S. Sankar, R. Siva, and S. Sharmila, "An
answers. effective E-learning framework model - a case study," 2009 7th Int.
Conf. on ICT and Knowledge Engineering, Bangkok, 2009, pp. 8-14.

314
 2022 IEEE 4th International Conference on Cybernetics, Cognition and Machine Learning Applications (ICCCMLA)

[3] L. Gaur, G. Singh, Jeyta and S. Kumar, “Google Analytics: A Tool to of MSR, 2018: 15th Int. Conf. on Mining Software Repositories,
make websites more Robust,” Proc. of the 2nd Int. Conf. on Inf. & Gothenburg, Sweden, 2018, pp.28–29.
Comm. Tech. for Competitive Strategies, Udaipur, India, 2016. [24] X. Xia, L. Bao, D. Lo, P. S. Kochhar, A. E. Hassan, and Z. Xing. “What
[4] U. Lee, Z. Liu and J. Cho, “Automatic identification of user goals in do developers search for on the web?,” Empirical Software
Web search,” In WWW '05, Proc. of the 14th Int. Conf. on World Wide Engineering, vol. 22(6), 2017, pp. 3149–3185.
Web, ACM, New York, USA, 2005, pp. 391–400. [25] https://towardsdatascience.com/random-forest-in-python-24d0893d51
[5] Andrei Broder, A Taxonomy of Web Search. SIGIR Forum, vol. 36(2), c0
2002, pp.3–10. [26] T. Muthuprasad, S. Aiswarya, K. S. Aditya, and G. K. Jha, “Students’
[6] P. Sondhi, M. Sharma, P. Kolari and C. X. Zhai, “A Taxonomy of perception and preference for online education in India during COVID
Queries for E-Commerce Search,” In SIGIR, 2018, pp. 1245-1248. -19 pandemic,” Soc Sci Human Open, vol. 3(1), 201, pp. 100-101.
[7] M. Tsagkias, T.H. King, S. Kallumadi, V. Murdock and M. de Rijke, [27] M. Al-Emran, I. Arpaci, and S. A. Salloum, “An empirical examination
“Challenges and Research Opportunities in eCommerce Search and of continuous intention to use m-learning: An integrated model,” Educ.
Recommendations,” SIGIR Forum, June 2020, vol. 54(1), pp. 1-23. Inf Technol vol. 25, 2020, pp.2899–2918.
[8] J.A. Hyman, M.T. Moser, M.T. and L.N. Segala, “Electronic reading [28] B. A. Kumar, and S. S. Chand, “Mobile learning adoption: A
and digital library technologies: understanding learner expectation and systematic review,” Educ. and Inf. Techs, vol. 24(1), 2019, pp.471–
usage intent for mobile learning,” Educ. Tech. Res. and Devept., vol. 487.
62(1), 2022, pp. 35-52. [29] N. Xi, C. Li and X. Yu, “Predictive Analytics of E-Commerce Search
[9] A. Kaushik, “A. Web analytics: An hour a day,” Indianapolis, Ind: Behavior for Conversion,”, Twenty-third Americas Conference on Inf.
Sybex, 2007. Sys., Boston, AMCIS 2017.
[10] M. Sleeper, S. Consolvo, and J. Staddon, “Exploring the benefits and [30] U. Kanjilal, and P.Kaul, “The journey of SWAYAM: India MOOCs
uses of web analytics tools for non-transactional websites,” In Proc. of initiative,” http://oasis.col.org/handle/11599/2592, 2016.
the 2014 Conf. on Designing interactive systems, ACM, NY, USA, [31] M. Singh, S.O. Adebayo, M. Saini, “Indian government E-learning
2014, pp.681–684. initiatives in response to COVID-19 crisis: A case study on online
[11] A. Phippen, L. Sheppard and S. Furnell, “A practical evaluation of web learning in Indian higher education system,” Educ. Inf. Technol. Vol.
analytics,” Internet Research, vol. 14(4), 2004, pp. 284-293. 26, 2021, pp.7569–7607.
[12] Gedrimiene, A. Silvola, J. Pursiainen, J. Rusanen and H. Muukkonen, [32] R. Bordoloi, P. Das, and K. Das, “Life long learning opportunities
“Learning analytics in education: literature review and case examples through MOOCs in India,” Asian Association of Open Universities J.,
from vocational education,” Scand. J. Educ. Res., vol. 64(7), 2020, pp. vol. 15(1), 2020, pp.83-95.
1105–1119. [33] M. C. Mendez, D. Shah, and S. P. Patra, "Exploring SWAYAM: How
[13] E. de Sousa, B. Alexandre, R. Ferreira Mello, T. Pontual Falcão, B. India is making MOOCs an integral part of its education system," 2019
Vesin and D. Gašević, “Applications of learning analytics in high IEEE Learning with MOOCS (LWMOOCS), 2019, pp. 64-68.
schools: a systematic literature review,” frontiers in AI, vol. 4, 2021, [34] A. Figueroa and J. Atkinson, "Ensembling classifiers for detecting user
pp 1-23. intentions behind web queries," in IEEE Internet Computing, vol. 20,
[14] W. Admiraal, J. Vermeulen and J. Bulterman-Bos, “Teaching with no. 2, Mar.-Apr. 2016, pp. 8-16.
learning analytics: how to connect computer-based assessment data [35] A. Strzelecki, and P. Rutecka, "Direct answers in google search
with classroom instruction?,” Technology, Pedagogy and Education, results," in IEEE Access, vol. 8, 2020, pp. 103642-103654.
vol. 29 (5), 2020, pp. 577–591.
[36] Welch, Tony Temitope, and Alade Nichol, Lynn. “Using the unified
[15] A. Wiggins, “Data-driven design: Using Web analytics to validate theory of acceptance and use of technology (Utaut) model to determine
heuristics,” Bulletin of the American Society for Inf. Sci. and Tech., vol. factors affecting mobile learning adoption in the workplace: A study of
33(5), 2007, pp. 20–24. the science museum group,” Iadis Int. J. On Comput Sci. And Inf.
[16] H. Luo, S. Rocco and C. Schaad, “Using google analytics to understand Systems, vol. 15, pp. 85-98.
online learning: a case study of a graduate-level online course,” Int. [37] R.Bonacin, J.C. Dos Reis, E.M. Perciani, and O. Nabuco, “Exploring
Conf. of Educ. Innovation through Tech. (EITT), 2015, pp. 264-268. intentions on electronic health records retrieval studies with
[17] R. Sentence, “Google Analytics: a guide to confusing terms,” 2016, collaborative scenarios,” Ingénierie des Systèmes d’Inf., vol. 23, no. 2,
https://www.searchenginewatch.com/2016/04/11/google-analytics-a- 2018, pp. 111-135.
guide-to-confusing-terms/
[18] H. Ahmed, T. A. Jilani, W. Haider, M.A. Abbasi, and S. Kamran,
“Establishing standard rules for choosing best KPIs for an e-commerce
business based on google analytics and machine learning technique,”
Int. J. Adv. Comput. Sci., 2017, pp.12–24.
[19] S. Asim, D. Jacob, N.M. Nawi, H. Mahdin, and M.E. Saputri, “The new
trend for search engine optimization, tools and techniques,” Indonesian
J. of Electrical Engineering and Comput. Sci., vol. 18(3), 2020,
pp.1568-1583.
[20] N. Akpinar, A. Ramdas, and U. Acar, “Analyzing student strategies in
blended courses using clickstream data,” 2020, ArXiv,
abs/2006.00421.
[21] D. A. Filvà, M. J. C. Guerrero, and M. A. Forment, "Google analytics
for time behavior measurement in Moodle," 2014 9th Iberian Conf. on
Inf. Sys. and Techs. (CISTI), 2014, pp. 1-6.
[22] M. Komenda, V. Bulhart, M. Karolyi, J. Jarkovský, J. Mužík, O.
Májek, L. Šnajdrová, P. Růžičková, J. Rážová, R. Prymula, B.
Macková, P. Březovský, J. Marounek J, and V. Černý, L. Dušek,
“Complex reporting of the COVID-19 epidemic in the Czech Republic:
Use of an interactive web-based App in practice,” J Med Internet Res,
vol. 22(5), 2020.
[23] M. M. Rahman, J. Barson, S. Paul, J. Kayani, F.A. Lois, S.F. Quezada,
C. Parnin, K. T. Stolee, and Ray Baishakhi, “Evaluating how
developers use general purpose web-search for code retrieval,” In Pro.

View publication stats 315