COVID-19 Infodemic in Malaysia: Conceptualizing Fake News for Detection

Table 3

Summarized procedures for data preparation.

Index	Procedure
1	The columns that require for future analysis are extracted
2	The presence of null or missing values of the extracted data are examined
3	Columns “Tajuk” and “Keterangan” are combined for future analysis
4	The news distribution is plotted based on the categories of the news
5	Column “label” is added, and the news is labelled into “real” and “fake”
6	The balance of data is checked
7	The data are preprocessed by natural language processing (NLP)
8	Frequency distribution of 30 most common token of words and word clouds for COVID-19 related fake news are generated
9	Feature extraction of the data is generated by term frequency-inverse document frequency (TF-IDF) with bigram
10	The data are split into 70% train data and 30% test data
11	SMOTE oversampling technique is employed to solve the data imbalanced problem