**4. Results**

Results of the proposed approach for the identification of factors affecting the health of students in COVID-19 will be discussed in detail in this section. Figure 4 explains the proposed method in detail with results. The results show that the dataset of a feature vector of 16 features is balanced through applying SMOTE technique. The health of students is taken as a target feature, and the Turf feature selection technique is utilized to detect the factors influencing the health of students. Different classification algorithms are applied to the selected feature datasets of student's health during COVID-19. The performance of the suggested method was assessed using accuracy, precision, recall, and f-measure assessment metrics.

**Figure 4.** Some feature selection process of factors affecting the student's health during COVID-19.

The result s shows that the student who utilized their time during lockdown period in COVID-19 in different activities remain healthy. Utilization of time appears as the main factor affecting the health of students. The academic organization may keep that factor in front and must plan activities, guide, and motivate students to participate in some indoor actives in such a way that maintains their health. Emotional attachment of students with family members also affects the health of students, as the fear of any family loss due to COVID-19 affects the health of students. Moreover, change in the weight of students during COVID-19 also affects the health of students. Figure 5 presents the results of four classifiers, GBDT, RF, SVM, and NN, on students COVID-19 dataset, whereas the accuracy describes the number of healthy students correctly classified by proposed work over a total number of students. Results show that a Neural network (NN) outperforms other existing classification algorithms in terms of accuracy. However, GBDT also performs well

on students COVID-19 dataset and showed around 87% of accuracy. Equation (7) presents the accuracy formula for the student COVID-19 dataset.

*Accuracy* <sup>=</sup> *Number o f studentscorrecly classi fied Total number o f students* (7)

Figure 6 presents the performance evaluation of the proposed work in terms of precision, whereas precision calculates the number of healthy students in the COVID-19 student dataset correctly classified by proposed work divided by the total number of healthy students in the COVID-19 dataset, classified by the proposed approach. Results show that neural network performs better than other classification algorithms. Equation (8) presents the formula of precision for calculation precision of proposed approach on student COVID-19 dataset.

**Figure 6.** Comparison of precision of proposed COVID-19 approach.

**Figure 5.** Comparison of accuracy of proposed COVID-19 approach.

The results in Figure 7 show the performance evaluation of the proposed work in terms of recall. The recall is the calculation of a total number of healthy students in the COVID-19 student dataset classified by the proposed approach divided by the total number of healthy students in the COVID-19 student dataset. The results show that the GBDT classifier outperforms other classifiers in recall performance evaluation measures. Furthermore, RF and NN also show better performance. Equation (9) presents the formula for calculating the recall for evaluating proposed approach on students COVID-19 students.

**Figure 7.** This Comparison of recall of proposed COVID-19 approach.

This is example 2 of an equation:

$$Recall = \frac{\text{Number of healthy studentsclassified by the proposed approach}}{\text{Total number of healthy students}} \tag{9}$$

Figure 8 presents a comparison of the performance of four classifiers in terms Fmeasure performance evaluation measure, whereas the f-measure of the proposed approach considers precision and recall both, presented already in Equation (6). Results show that GBDT and NN give better performance on the proposed work on the COVID-19 student dataset in terms of F-measure.

**Figure 8.** This Comparison of f-measure of proposed COVID-19 approach.
