Page 146 - IJEEE-2022-Vol18-ISSUE-1

142 | Atiyah & Thalij

TABLE I D. Results of Classification

DESCRIPTION OF DATASET The performance of ML-algorithms that used in classifying
COVID-19 disease is evaluated. Accuracy, specificity,
Feature Kind Prescribing accuracy, sensitivity, negative-prevalence and positive-
prevalence, ROC-AUC-Score, mislabeling and execution
PATIENT_VISIT_ int_64 identifier of patient time measures were used for these algorithms. A first model
comprising four algorithms was conducted, and then a
IDENTIFIER that visited the second classification model was made, which also included
four algorithms, on the same global dataset to test the largest
hospital number of algorithms to choose the best. Table 5 shows the
performance of the algorithms used on the test set, while
GENDER int_64 gender of patients Tables 6 show the performance of the algorithms in terms of
execution time and misclassification. After making a
AGE_PERCENTIL Object percentile of ages the comparison of the algorithms in Tables 5 and 6, SGD the best
AGE_ABOVE65 int_64 patients. performance among the classification algorithms used when
we chose classification is based on the labeling of the ICU.
the ages of patients While KNN was the worst performance among the
that above 65 years. algorithms in those works.

DISEASE_ float_64 six sets of the diseases

GROUPING that have available

attributes of the

nameless information TABLE V
COMPARISN OF ALGORITHMS PERFORMANCE
.........

RESPIRATORY float_64 the available measurements SGD DT RF XGB SVM NB LR KNN
_RATE_ attributes around the unit
DIFF_REL respiratory average
relative -diff

the available Accuracy 99.61 99 98.4 98.4 97 96.6 94.8 88.4
attributes around the
TEMPERATURE float_64 temperature Sensitivity 100 100 95.4 90 85 90.1 80.3 45
_DIFF -REL relative_diff

the accessible Specificity 97.43 94.8 94.9 97.4 94.9 89.7 97.4 100
attributes around the
OXYGEN float_64 oxygen sated relative Precision 95.23 90.9 90.5 94.7 89.5 81.8 93.3 100
SATURATION _diff 92.5 94.6 86.4 78
DIFF -REL Negative 91.5 96.6 94.8 81.4
prevalence 100 100 97.4 95 89.9 89.9 83.7 72.5
The windows has five
Positive
WINDOWS object kinds of sets everyone 99.61 96.6 98.4 94.9
containing hours for
acceptance prevalence
Roc-auc-
reply features (0
98.71 97.4 94.9 93.7
score

ICU int_64 represent not need to TABLE VI
ICU and 1 represent COMPARISN OF EXECUTION TIME AND

need to ICU) MISLABELING

TABLE II Measurements SGD DT RF XGB SVM NB LR KNN
THE PATIENTS’ TOTAL. unit

Patients Total after pre-processing 293

need to ICU 105 Execution 0.01 0.02 0.2 0.18 0.03 0.04 0.7 0.01
Time
not need to ICU 188

TABLE III mislabeling 0.39 1 1.63 1.63 3 3.43 5.18 11.6
DISSEMINATION OF AGE FOR TOTAL infections.

Age Dissemination

infections under of 65age 172 V. CONCLUSIONS

infections over of 65 age 121 The COVID-19 epidemic a become a major disquieted for
countries worldwide and has affected millions of people, has
TABLE IV become represented a major economic and social challenge,
DISSEMINATION OF AGE FOR INFECTIONS IN ICU and poses a serious threat to public health, so there is
required to classify COVID-19 dataset to forecast the
Age Dissemination 60 number of injuries when disease outbreak to avoid a collapse
Infections over of 65 age: 45 the system healthy. Moreover, in cases of COVID-19 with
Infections under of 65 age: varying severity grades, some require ICU. Therefore, it is
necessary to forecast of injuries number that require ICU to