The amalgamation of XGBoost and SMOTE for sentiment analysis pedulilindung application reviews

The increase in daily Covid-19 cases in Indonesia was caused by the new year’s holiday and the cancellation of the Level 3 Restriction on Community Activities, thereby increasing community mobility during holidays. To overcome this, the government optimizes the use of the PeduliLindung application....

Full description

Saved in:
Bibliographic Details
Main Authors: Singgih, Malecita Nur Atala, Utari, Dina Tri
Format: Conference Proceeding
Language:English
Subjects:
Accuracy
Classification
Data analysis
Data mining
Sentiment analysis
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The increase in daily Covid-19 cases in Indonesia was caused by the new year’s holiday and the cancellation of the Level 3 Restriction on Community Activities, thereby increasing community mobility during holidays. To overcome this, the government optimizes the use of the PeduliLindung application. Applications can be downloaded via Google Play and can get reviews from users. This feature can view reviews containing positive or negative opinions or comments. Based on the PeduliLindung application review data, sentiment analysis can determine how the community responds to the PeduliLindung application. The classification methods used are Extreme Gradient Boosting and Synthetic Minority Over-Sampling Technique to overcome data class imbalances. This analysis shows an increase in F1 score, accuracy, and AUC compared to before data balancing, which are F1 score of 75.00%, an accuracy rate of 95.43%, and an AUC of 0.84. These results mean that the model’s accuracy in classifying and predicting positive and negative reviews can be said to be good.
ISSN:0094-243X
1551-7616
DOI:10.1063/5.0204739