Contributed By: MAIMOONA JASMINE
Background: I am Ar. Maimoona Jasmine; I am an Architect who is into practice as well as education. I was working as an Assistant Professor of Architecture at MGR Educational and Research Institute University in Chennai. My husband is a QMS consultant with Cognizant, due to his influence and the scope, and flexibility offered by the IT industry; I was planning to switch over to IT related career path. Hence chose to study the PGPDSBA course offered by Great Learning. Here the institution I worked for though renowned, is still very outdated and primitive in the means of handling data and let alone analyzing it. The process of storing data of various activities and the details pertaining to students and the management was digitalized very recently like a year or so. I find it an interesting platform to explore whatever was taught to me during the duration of the course and to visualize a few hypotheses that always existed in the education sector.
Problem Statement: Here, I wanted to check whether the following believed notions were true or not:
1. If the attendance of a student really matters when it comes to the final score obtained. If yes or not, what would be the reasons that actually determine the final scoring of the student?
2. There is a concept called Continuous Assessment Test (CAT) exams that are conducted twice or thrice every semester and the marks obtained in it are computed for the final internal score. These CAT exams test the consistency of the students. So, it is generally believed that the student scores in the CAT exams determine the final internal score.
3. Also, I wanted to create clusters on the available data and to see how to work with the students that fall under each category. And to check if the segmentation done on the data of two subjects would also be applicable to other subjects. So as to check if a poor-performing students cluster has poor marks in all the subjects or are there exceptions.
These are some of the tests I wanted to conduct on the available dataset, containing final internal marks of the 2020 – 2025 batch in their first-year 2 nd semester in 2 subjects which are highly varying in nature. And wanted to visualize the results and share the insights with the institution. This batch is currently in their 3 rd year. So, the clusters thus formed still hold good or not could be checked. And these insights derived from the attendance, CAT and result hypothesis could be used as a base to coin some useful strategies to engage the students in a better manner.
Analysis: Some other analyses I would like to do for the institution are:
1. Understanding the type of students that are getting admitted into the institution – their feedback, HSC scores, their demographics and building a predictive model for finding if a student who has got the application form will join the institute or not.
2. And based on data of the students currently enrolled and those who have graduated, find the clusters which would be the target communities or groups likely to join the institute and do dedicate marketing campaign for those specific groups.
3. Also, to check if their current advertisement levels have any impact on the admissions received, and which would be a better means of advertisements based on the data.
4. Faculty performance based on different attributes like results of students, feedback from students, research papers and journals published and the involvement of the faculty in the other extra-curricular activities in and for the institute.
5. Does the internal score and the final marks obtained from the university final examinations are they correlated or not? If not then where is the difference?
Strategy and Objective: The idea here is to indulge the higher management in visualizing whether the notions and the basic ideologies they have conceived are true or not. The same analysis could be performed in any random samples and the results may vary. But understanding what the data communicate might be the key to taking decisions with the data as proof than just speculating on the results and deciding based on notions which is generally the case as the institution is very traditional in its approach and is just learning to embrace technology.
Tools and Techniques Used: I have used basic Exploratory Data Analysis, ANOVA to check the hypothesis, and Hierarchical and K-means clustering for obtaining student segregation. Have used Python for data visualizations. I have been able to find the following after the analysis:
1. The attendance and the marks obtained have very less to no correlation, also that attendance has no impact on the marks obtained by the student as it is generally believed. This is the case in both
of the subjects in the dataset. But as to what is the influencing factor for scoring is a vast subject that requires further analysis and a bigger dataset.
2. The performance in the CAT does have a positive impact on the final internals, but not as much as was speculated. In one of two subjects, the CAT marks had no impact at all on the final internal. This could be because of various reasons.
3. The clustering does a good job of dividing the students into high-performing, medium-performing and poor-performing students. And the same performance is seen in both the subjects tested. Hence it can be concluded that the clustering has divided the students into proper groups. It can be noted in the clusters that high-performing students have good attendance and medium-performing students have lesser attendance and poor students have bad attendance. But the previous ANOVA test shows attendance has only a 67% impact on the scores.
Solutions Proposed: The solution recommended to solve the problem:
1. To address each cluster of students separately and to find what the issue is with the poor-performing students- checking on their demographics data and individual counselling sessions and to involve the good and medium students on various extra courses as per interest or exposing them to competitions to help them gain more experience.
2. Also, as for attendance though it has not had much of an impact on the scores. Lesser impact of attendance means the students are able to grasp the subject irrespective of their absence, hence it is to be taken into account that the study material and the content of the sessions reach the students so they can self-learn.
3. CAT scores do have an impact so the students are to be made aware of that and should be recommended so as to not miss out on any assessments and an alternative method must be designed for students who performed poorly or were absent due to valid reasons to help them retake on the assessment.
Impact: The impact of the study on the organization was understanding the important factors in a better manner with the visualizations to support the claims. And the same is to be done on a larger, random and diverse sample of the data to see if the patterns observed are the same in all the samples, so a progressive decision could be arrived at the methodology to be approached for the benefit of the students and the institution. This exercise has shown me the scope of data analytics and its usage in every field both at micro and macro levels. Analytics of the data can lead to a positive impact not only in the financial aspects but also on the well-being of the masses and social upbringing too. Also, the visualizations help everyone to understand the insights and the inferences better. It provides a different perspective on the same problem in a different aspect which opens one’s mind for better decision-making.
