Market Analysis of Key Manufacturing Segments Using News Data

This paper examines the use and viability of unstructured data in forecasting the real economy in order to quickly understand the current situation and trends in the real economy in light of growing uncertainty both domestically and externally. To extract an index utilizing news data, the methodology of Thorsrud (2016) was used to develop two approaches. The first approach involves conducting topic analysis to extract topics and then employing sentiment analysis for each topic and calculating a simple index. The second approach involves creating a comprehensive score by combining the sentiment score and topic score to create a sort of composite index. To prove the utility of news data-based indices, the total population was set to match the number of topics that had the maximum number of word groups of all of the news data from the six segments of the manufacturing industry for the given period. That number was then reduced based on topics that met a certain standard, and the correlation between the extracted indices and real industry indices (such as an increase in the segment-based manufacturing index) was examined. In conclusion, the numbers calculated by analyzing the news for each segment of the manufacturing industry showed a similar trend to that of the real economy, showing that data extracted in a quantitative manner from unstructured data such as news articles prove to be significantly useful in understanding trends in the real economy.