Analyze Corporate Anti-Corruption Disclosure with Feature Selection

Corruption creates problem for a country, including Indonesia. Indonesian government put their effort to eliminate corruption. Not only the government, but companies also have their part in corruption elimination with anti-corruption disclosure policy. For some time, the research tries to find the significant variables that influence the policy. However, most of the studies use a statistical approach only. This study offers an alternative method called feature selection that is part of machine learning. The data set in this research has two sources. The first one is sourced from the annual report of companies listed on the Indonesia Stock Exchange from 2017 to 2019. The second source is the membership data of the companies in the United Nations Global Compact (UNGC). Experiment shows that feature selection methods can offer alternatives variables selection. Based on prediction performance in machine learning, feature selection methods may choose the factors that are important in company anti-corruption disclosure policy better than statistical analysis.