Analysis, Prediction and Maintenance of Teaching learning process based on empathize Students’ View of attending Online/Regular Class

Learning models have been widely used in predicting diseases, disorders, behaviour aspects in human beings etc. The current research gives an analytical study on predicting University students’ behaviour with various Machine Learning approaches. Research shows that Machine Learning approaches outwit the strategies especially in Student behaviour analysis. An analytical study on various learning approaches and its application in Behaviour Analysis is vividly presented in the paper. The study would give an understanding on how various learning approaches could be applied in Student Behaviour Analysis that includes academic performance, behavioural study with reference to courses, Online teaching modes etc. The paper also encompasses comparison with various Machine Learning approaches in student behavioural prediction.


Introduction
Deep Learning, an emerging research area has its usage in wide variety of applications starting from Health care to many other including Education. It is a subset in Machine learning in which model mimics the brain through the implementation of neural networks. Deep Learning encompasses Deep Neural networks (DNN), Recurrent Neural Networks (RNN), Convolutional Neural Networks (CNN) and Q Learning. Recurrent Neural Networks is a network which is used for handling time series data. It is a type of Artificial Neural Network where each unit is connected in a sequence and a directed graph is formed. The connections between each unit are in the form of a sequence. Each element performs the same task hence it is a Recurrent Neural network. Behaviour analysis on students is a general study done to check their academic performance. Automation of student activities in a University analyses and predicts students' performance and makes the relationship between them and academic activities [1]. Behavioural study also is the study of the mood with which an individual reacts with reference to content [2]. Similarly, online courses have led to drop-out of candidates through literature survey it is very evident [3]. Classroom atmosphere is also responsible for students' behaviour Vithya Ganesan et al. 2 which is predicted through their behaviour signals [4]. The situation during COVID 19 is that all classes which were happening at Universities have been made online which bridges several gaps between direct classes and online courses [5].
Current analysis of the behaviour is of utmost importance in today's scenario for several reasons and hence the paper focuses on the same Section II represents Student Behaviour Analysis Study, Section II elucidates the Deep Learning application in Behavioural Analysis, Section III contains results and discussion and finally concludes with Section IV Conclusion and future work.

Student Behavioural Analysis Study
Student behaviour analysis before and during COVID 19 being the proposed work helps to identify the attention levels of students. Here an exhaustive literature survey on student behaviour analysis has been carried out.
Famram Ali Khan, et.al elucidated the understanding of teachers on students in an online learning environment with reference to their affective states, learning styles, student learning preferences. [6] Xiang et.al described that students' academic performance is related to other behaviour factors especially the way internet is being used. [7] Hafed et.al analysed that students' behaviour could be identified with their involvement in social media like Facebook in which data was collected during the learning session. [8] Sujit Kumar Gupta et.al elucidated the study of students' behaviour through a face detection method for analysing the content of videos. [9] Yang et.al proposed that with the students' in-classroom behaviour helps in evaluating the efficiency of teaching which teachers do at class. The evaluation is done by detecting students' faces, head raising or downing faces, the head orientations of the teacher, the extraction of the audio features of the teachers' speech. [10].

Machine Learning in Behavioural Analysis
Machine Learning is widely used in the study of students' behavioural analysis. The analysis done on students includes models which can be applied on secondary school students to seek admission in Universities where the relationship between cognitive and psychological variables are taken into consideration for depicting the performance of secondary school students in academics by using Artificial Neural networks approach. [11], [14] Analysis could be done on various factors that would influence academic performance like marks in the degrees obtained, home environment, Study learning habits, hardworking nature, academic interaction, stress level, level of friendship with group members and level of mental relaxation using artificial neural networks. [12], [13].
So, Feature selection method with more optimal feature subset which gives higher accuracy in prediction is required in analysis [15] [16] specifically to identify lack of sitting tolerance, lack of attention, learning disability [17] [18].

Pre-processing
Data Pre-processing is a process of making a raw data into an understandable format. There are various ways in which data can be pre-processed. Five step process to analysis the Teaching learning process either online or Regular classes 1. Empathize students thought, understand their learning habit, and discover students hidden talents. Analysis, Prediction and Maintenance of Teaching learning process based on empathize Students' View of attending Online/Regular Class 3 The following sections explain the design process in teaching learning methodology.

Empathize students thought, understand
their learning habit, and discover students hidden talents.
It is by observe, interview and collecting questionnaires from students. Data is collected by the following. I. Flooding the questionnaires II. Simple observation III. Shadowing

I. Flooding the questionnaires
The following are the questionnaire asked in the

II.
Simple observation: Become one of them or work alongside with leaners and mechanical observation like eye tracking to improve the teaching learning process.

III. Shadowing:
Watch or keep track on leaners learning habits.

Online class strategies of Teaching learning process needs:
Online consciousness live status, Recording of Live Class, creating a timetable & conducting Multiple session concurrently, Administration of Quality of classes with various reports for management, Muting student's mic & allowing one by one, Raising hand

Regular Class strategies of Teaching learning process needs:
Continuous Evaluation, Home Assignment, surprise quiz, Tutorial, Hackathon, Field Study review, prototype review, Group discussion, industry connection program, global certification, promote to poster presentation, Leader board ranking for global challenges, Capstone project, paper publication, presentation, exercise Flipped Class: It is a hybrid class, combination of online and regular class. Vithya Ganesan et al. 4 Regular Class: Promote poster presentations, Leader board ranking for global challenges, Capstone project, paper publication, presentation, exercise. Figure 2. Data analysis of student activity Table 1 explains the feedback analysis report. Figure 2 shows the data analysis of student activity. Figure 3. Shows the Multicollinearity of student data analysis.            Table 2. Classification report of the model

Conclusion
The demographic shows the insight of students thought, feelings, likes, Attitude, fear, habit, hobby, influence, and constraints. Machine Learning classifiers with a highperformance accuracy for training and evaluation.
From the set of data, most of them likes regular classroom is their preferred Mode of Study for the year 2021. It is shown by the real data analysis and waiting for to attend regular classes.