Leveraging Textual Data in Nowcasting Malaysia’s Gross Domestic Product
64th ISI World Statistics Congress - Ottawa, Canada
Format: CPS Abstract
Keywords: gdp, machine learning, news sentiment, nowcasting, sentiment analysis
Session: CPS 73 - Finance and business statistics V
Wednesday 19 July 8:30 a.m. - 9:40 a.m. (Canada/Eastern)
Generally, nowcasting economic indicators such as Gross Domestic Product (GDP) is often making use of structured data including real, financial and survey indicators. Recent research focused on how textual data from news and social media have been utilized in improving nowcasting models. There are two methods used to perform text analysis which are machine learning methods and lexicon-based methods. In this paper, we consider online news sentiment analysis as an additional data source into the recent work of nowcasting Malaysia’s GDP using Machine Learning. This paper performed both machine learning and Lexicon-based sentiment analysis to capture news sentiment on economic activities, i.e. Valence Aware Dictionary and Sentiment Reasoner (VADER), Maximum Entropy and SparkNLP. XGboost with sentiment analysis using SparkNLP found to perform better in nowcasting Malaysia's GDP.