Currently, society is facing a pandemic known as coronavirus disease 2019 (COVID-2019). COVID-19 has exposed existing deficiencies in all sectors of society, with education being one of the sectors affected by the pandemic [1
]. To face it, educational institutions have added information and communication technologies (ICT) to their educational and administrative processes [2
]. This does not mean that this union between ICT and education did not exist before, it is simply that because of the pandemic they have a more active role in each educational process. An example of this is that ICT are generally applied in administrative management processes, such as tuition payments, human resources systems, payroll payments, enrollment systems, etc. With COVID-19 spreading throughout the world and without encountering any resistance that contains its high contagion capacity [3
], society has been forced to implement lengthy isolation or quarantines. To continue with their activities, educational institutions, especially universities, have integrated videoconferencing systems into their infrastructures in order to generate synchronous meetings to replace face-to-face classes.
By replacing face-to-face classes with synchronous meetings, students and their learning are directly affected [4
]. For this reason, universities are studying educational projects that allow them to change their educational model, focusing on meeting the new needs presented by students. Most of the educational projects that are analyzed by universities are not new, however, they have only been applied by some universities. Among these models are virtual education, online education, and coeducation [5
]. These models have already been developed and there is a great deal of information on how they work. In these, the center of learning is the student, and they use resources and interactive activities that improve student understanding [6
]. Once the student is in charge of learning it, the teacher takes the role of tutor who is in charge of clearing up the doubts that the student brings with him.
The pandemic and the measures that countries have taken to try to control it have changed society [7
]. The use of ICT has allowed the continuity of most activities, which in the past focused on face-to-face attendance. With this reality, the future of education is definable, where its entire environment is modified and aligned to meet the needs of students. Mixed education is the alternative that is closest to the future, since having the best characteristics of face-to-face education and online education marks an advantage that adapts to the student, even with the integration of new technologies or emerging technologies [4
]. It is possible to speak of a personalized education.
However, for the blended education model to be scalable and adaptable to the student, it needs to integrate ICT specifically in two important aspects. The first aspect is data analysis that allows the monitoring of students by simulating the areas of academic monitoring of face-to-face education. Data analysis has the advantage of obtaining information on all the activities that the student performs, regardless of the educational modality [8
]. If necessary, the analysis includes psychosocial data to determine problems related to the student’s emotional state, quickly identifying problems in order to generate efficient solutions so as not to impair learning. Another aspect is decision-making; ICT have to provide systems that are in charge of notifying and giving quick solutions to students, always considering their learning [9
This work takes as a reference the use of a data analysis architecture that allows identifying the needs of university students, and the results are analyzed to generate improvements in their learning [10
]. For this, artificial intelligence (AI) techniques are used that take the knowledge generated from data analysis and interact with the student and with the areas in charge of academic monitoring [11
]. This work is divided into the following sections that have been considered key to reach the proposed objectives. Section 2
defines the materials and method; Section 3
presents the results obtained from the analysis; Section 4
presents the discussion of the results obtained with the proposal for improvement in the educational modality to improve learning; Section 5
presents the conclusions found in the development of the work.
The environment where the system is implemented is a university in Ecuador, which has a face-to-face academic proposal. To continue their activities during the pandemic, they have integrated a platform into their infrastructure that allows them to run synchronous meetings. The student population is close to 10,000 students divided into the matrix located in the city of Quito and two branches located in the main cities of the country. This work has been applied to the matrix that contains around 4000 students belonging to five different faculties. The analysis was carried out on first semester students in a remote mode as a result of the pandemic. The objective of conducting the analysis in this group is to determine if there are possible cases of attrition. In addition, it seeks to determine the state of learning in the group. The number of students is 18 people, this course belongs to the career of computer engineering, so the use of ICT in most of its activities is frequent, which ensures a large volume of data.
The expert system has been developed with the intention of being transparent in the interaction with the student. This means that it has been developed as an additional module of the LMS, this being the system that the student uses for most academic activities. Therefore, the student will not have to enter an additional system to observe her own performance. The system, at the moment the student accesses the LMS accesses the results of the data analysis in search of the student′s academic performance. If there is any pending activity, the system sends notifications through emails to the student and presents it on the LMS calendar. If there is any grade lower than 6/10, which is the average established by the university as a minimum grade that the student must meet to pass the different subjects, it triggers an event in the expert system that is responsible for starting the process for the recommendation of activities.
In the first stage, students with problems in academic performance are identified. Table 1
shows the results of the analysis, for which academic data sources, a synchronous meeting system, and financial sources were integrated. The relationship of the data obtained is as follows: the academic sources provide personal information of the students that, to present them, an identifier has simply been placed for the proper handling of the information. This system contains the final grades of the students, by university policy the students who do not reach the score of 6 points repeat the subject, which in this study is Programming I. The third column is information that the system has crossed with synchronous meeting system data. This system records the connection time that the student has. However, it is not simply the time a student starts the session and when it ends, but the system has the ability to detect if a student shows activity during the session. This is done through the detection of events that it has with the peripherals from your computer. For example, the camera detects the presence of the student, the software detects the activities that the student has, and issues reports on this interaction. The keyboard and microphone are used to detect the dedication of the student while in the synchronous session. The academic period is composed of 16 weeks of which 14 are effective weeks of classes with three hours a week assigned to the subject. Two weeks are used by the teacher to carry out comprehensive evaluations to students. Therefore, a student who effectively completes his classes through synchronous meetings will have 42 effective hours.
In the fourth column, the system adds the data from the financial system, and these data are obtained through surveys conducted with students. The objective of the surveys is to determine if there are financial problems for students or their families to generate tuition payments, given that the university is a model of private education that has its own financing. A weight was added to the answer to present them in the results with the information cross-referenced by the Hadoop framework. The survey presents several questions on various topics of interest to the university. However, the following question has been included in the analysis with its corresponding weights:
Due to the pandemic, you have financial problems to cover tuition costs:
Very high: 1.00
From the results obtained there is information that is easy to relate. For example, the majority of students who meet or exceed the average of 6 and pass the subject exceed 33 h of effectiveness in class. There is a case that must be analyzed because it exceeds the average of 6, but its effectiveness in classes reaches 17 h of the 42 required. There are six students who do not meet the requirements to pass the course and the study continues with these students. Table 2
shows the cases that have been considered for a granular analysis to determine the causes of their poor performance. The data presented in this table is easy to understand thanks to the Hadoop tools used. As a first analysis of these results, it can be determined that grades have a direct relationship with the number of effective class hours of students. One reason for this is the change from face-to-face to remote modality, which was so rapid that, in the change, the institutions maintained a face-to-face educational method. In other words, the resources, activities, and all that correspond to the material used for the development of learning were not modified. Even something as important as the syllabus of the subjects that are the main resource that helps students define the rules and norms that they must follow in each subject was maintained. The problem arises in that this method works when the main actor of learning is the teacher, who through face-to-face classes becomes the learning sensor. He is the one who identifies the existing problems within the classroom, even with the interaction with the student he can become the motivator of learning, defining what the student should learn and how he should do it.
When trying to replace face-to-face classes with synchronous meetings, a problem is generated that beyond learning is based on the motivation that the student feels through these tools. A proof of this is that the students presented in the table have a low number of effective hours of classes in relation to those who passed the course. However, it should be noted that in this group the problems with tuition payments are more accentuated. This is a factor that undoubtedly affects students, even creates a negative learning environment; there are cases where students, knowing that their economic projections do not cover the payment of the next tuition, abandon the subjects they are studying. Specifically, in these results we have three students with high problems in their economic situation. If to this factor we add the low number of hours of classes in the period, a low performance is imminent.
The next step in the analysis is to determine the cause of the problem that pushes students to present low academic performance. To identify the problem for analysis, a new data source is integrated, which is the LMS DB. In this DB the information about the activities that the student carries out are recorded. Table 3
shows the details of the activities that the student developed in a normal course of classes. The activities are proposed by the teacher and respond to the needs and issues that the teacher decides to evaluate. In the Programming I course, students must comply with two types of activities, autonomous activities are programming exercises raised on the topics that are developed during synchronous meetings.
The tasks are graded on 2.5 points and the first total takes the grade point average, then there are the continuous evaluation activities, reflected as questionnaires. These activities evaluate the conceptual knowledge of the subject; they are quick questionnaires where students choose one option or several according to the question posed. The next column labeled “Total” takes the average of the questionnaires that have been raised during the academic period. The activities add up to five points and the rest are evaluated in exams. The column “Prac” refers to the practical test and the “Theor” column to the theoretical exam. The distribution of scores is three points for the practical evaluation and two points for the theoretical evaluation. The “Final Total” column corresponds to the sum of the activities and exams and is the one that determines whether a student passes the subject.
In the results presented in the table, the information is not clear, or it is difficult to understand what is happening in the analysis. Therefore, in accessing information, the Hadoop framework deploys several tools that allow interpreting this knowledge. In Figure 7
, the performance of the different activities per student is presented, in the student axis are the identifiers of the six students analyzed. In this way, it can be determined that students 5 and 8 have a better performance in the development of questionnaire-type activities where the conceptual knowledge of the subject is evaluated. What is directly aligned with the development of the theoretical exam, this relationship is the one that needs to be exploited with the recommendation system.
In the case of the student with identifier 13, this relationship is more significant, since in the development of practical activities the performance is better than that found in the questionnaires. In addition, the performance in the practical exam is aligned with the results obtained in the practical activities. Similarly, in the following cases, although there is little distance between practical activities and questionnaires, it is possible to determine which activity best aligns the student.
Next, the expert system takes the knowledge obtained by the data analysis framework and begins its processing to recommend improvements or activities. In Figure 8
, the block diagram representing each stage of its processing is presented. The system starts with an event that is generated when the student logs into the LMS. The system reviews the students’ performance; in the case of the six students analyzed, it finds that there are learning anomalies by not achieving an average equal to or greater than 6 points. Identified, the student takes the data of her performance and processes them as a person does. In the data, find what is the deficiency of each student and in which of the activities you find a greater problem for their learning. In the process, the system compares the activities carried out by the student, the grades obtained and according to these compares with its base of rules to determine the activity meets these needs. The relationship by which he reaches a conclusion is as follows:
Gets information from the fact base that is the results of the data analysis.
Compare with your rule base, where it is concluded, that if a student demonstrates greater effectiveness in the development of practical activities and problems in the development of questionnaires. The system searches its knowledge base for activities that have this practical approach, such as project development, challenge activities, etc.
If the student easily develops the questionnaires, it shows that conceptual knowledge is a means where the student′s knowledge can be exploited. Search the knowledge base for activities that meet this requirement and recommend the student to develop forums, discussion papers, concept maps, etc.
Once the processing has been determined and the activity to be recommended identified, the system presents the results to the student and the teacher. It processes this information and executes the action within the LMS. In addition, it sends notifications of the conclusion reached to the parts that interact in the learning. Once these occur, the system enters a waiting stage; this represents the time it takes the student to complete the activity and for it to be graded. Once these actions have been executed, the framework analyzes the data again and passes the information to the expert system where it is verified if there was an improvement in learning. If the student′s performance improves, the system takes this data and stores it in its knowledge base, which is useful for other cases with similar characteristics, and exits the process.
presents the results of two students where the system is evaluated. The students included are those with identifiers 5 and 8, in the previous analysis it was determined that these students have greater facility for the development of questionnaire-type activities. The expert system recommended the replacement of activities that are 100% practical with activities where students feel more comfortable in their development. With this recommendation, the student was asked in the new period the development of the activities indicated in the table and in its detail. When comparing the results of these students with those of the previous period, the improvement is significant. It is even demonstrable that, if these changes were made on time, students would hardly miss a subject, and they guarantee learning. In addition, it should be noted that by including new activities such as forums, concept maps, as well as research reports, the student improves in the development of practical exercises. This is attributed to a better understanding and application of the conceptual part in a practical environment.
The recommendation of the activities is carried out by the system by the formation of groups, as it was done in the evaluated case. However, you can also do them individually by customizing the learning.
Universities go through decisive moments, considering that education and the way it is run has changed. Well, the pandemic and ICT have shown that it can be carried out in remote environments. However, as in any organization, it is necessary to create scenarios where the application of technologies helps learning. In the case of universities, they have the advantage that they make use of many ICT tools, which improves the generation of data that can be processed in search of knowledge. From the results found in this work, we can emphasize certain key points, the first being the data analysis. Data analysis in universities is a common activity, but it is commonly applied to marketing management or financial decision-making. This work adopts a robust architecture in the analysis such as big data. The reason for this decision is due to the fact that education in the future will involve the use and integration of emerging technology. Many of these technologies present a great variety in the data, therefore, this work has been developed in such a way that it is scalable and adjusts to any situation generated in learning.
The next point is the recommendation system, in a remote education model it is common that there are several tasks where human management loses effectiveness. For this reason, AI through expert systems or recommendation systems are key to address the tasks where the efficiency and effectiveness of the systems is needed. According to the results, it has been determined that both the data analysis and the AI are efficient. However, it should be noted that being in a testing stage there are variables that were not considered. This does not mean that they were omitted from the analysis, it was simply established a controllable environment that can be conveyed in this document. Furthermore, the questions posed did not include a major inconvenience in data processing. Which is presented as an excellent projection in the use of Hadoop. With the data obtained and passed to the factual base of the expert system, it is possible to create a number of possible events that the expert can store in the knowledge base, speeding up the recommendation of activities.
The expert system has been designed as its own application aligned to the needs of the university. This guarantees that if there is a granular analysis, the system can provide feedback that allows its processing to be adjusted to the point of personalizing the recommendation of activities that is currently carried out through groups. The identification and classification of individuals in each group is carried out by the Hadoop framework. If in a very short time it is necessary to include a real-time analysis, the analysis architecture can integrate Apache Spark into the processing.
Several works focus on performing an analysis of educational data to identify the factors that influence the academic performance of students. However, it is necessary to consider that, in a comprehensive solution, the objective is to make decisions about the results obtained from the data analysis. Our proposal aims to improve learning and with it the educational model, for which a model of recommendation of academic activities is integrated into the data analysis. This model uses artificial intelligence to generate knowledge about the data obtained by having tools, such as big data, which are responsible for identifying patterns in student data and classifying them. Artificial intelligence identifies the academic deficiencies of each student and is able to recommend activities focused on students′ needs. Other works focus on the use of intelligent techniques that, through the use of statistical algorithms, identify the percentage of incidence of each factor in learning. These models, although they are effective, are not capable of handling large volumes of data, therefore, they are temporary solutions. Universities need to create scalable infrastructures that guarantee the operation of academic models in the short and long term. By integrating big data and artificial intelligence, a robust architecture is generated with the ability to work in various academic areas, improving the educational model and providing the scalability that is needed to create new educational projects.
With the results obtained from the evaluation of the proposed system, it is possible to highlight the ability to create an educational model based on the projections of each student and the detection of their weaknesses. To make adequate and effective decisions, however, it is necessary to emphasize that implementing a model such as the one proposed requires a consistent testing period, which can lead to a disadvantage. Since, at present and with the validity of the new normality, rapid and agile responses are required that provide solutions to the problems that universities are already facing.
The current situation in which society finds itself due to COVID-19 has changed the way in which it develops. The confinement to which society is subjected has forced the different sectors and areas to seek solutions based on the use of ICT to guarantee the continuity of their businesses or activities [49
]. Universities as well as other organizations have introduced new tools to their activities to continue with education. The tools that have been integrated in most cases are video conferencing platforms. However, the integration of these tools solved the problems that arose at the beginning of the pandemic. The main problem being the development of face-to-face classes which, in the traditional educational modality, is the main activity for the development of education [50
In the future, education will continue to evolve at the same speed as ICT, which is why all learning analytics work must focus on comprehensive solutions. This implies that data analysis will not be enough, systems that generate knowledge of the results of big data must necessarily be integrated. Adding value to data, although it has been a people-centered task, needs more speed and effectiveness. Systems that include AI are ideal for this type of application where the ability to solve a problem effectively allows even academic follow-ups and ICT become the ideal assistants in the academic development of students.
The lack of effectiveness in learning is an issue that is reflected in the dropout rates, as well as in the low academic effectiveness that is measured by the number of graduates in the different cohorts. These problems are accentuated in a future where it is considered that education will not be the same again [51
]. The integration with new technologies in the educational field will accentuate the proposal of mixed educational models, where learning necessarily focuses on the student and the university focuses its efforts on meeting the needs of the students, including in the execution of a personalized education.
The recommendation system, as well as the identification of the factors that affect academic performance, allow making decisions that contribute to the improvement of learning. The results obtained guarantee the scalability of the system, with which it is possible to add a greater volume of data [52
]. The recommendation of activities allow to significantly improve the grades and the learning of the students [53
]. These results undoubtedly make it possible to reduce the dropout and repetition rates that, due to the pandemic, have suffered a considerable increase in the implementation of a remote education model.
The integration of technologies in academic settings is another factor that is considered in this study and future works. Although this work includes big data and AI, there are technologies, such as the internet of things, blockchain, cloud computing, etc., that should be included in educational systems. The integration aims to improve all aspects and environments where an educational event takes place.