Umeå universitets logga

The context of this study is the use of news sentiment analysis to analyze the state of the energy sector in Ukraine during a full-scale war. The objective of the study is to use sentiment analysis based on the judgments of the large language model GPT-4 to determine the sentiment of news at different time intervals. The method used in the processing includes collecting data from a Ukrainian news site, preprocessing the data and applying the algorithms of the large language model to determine the sentiment level of the data. The results of the study include a set of news with assessed sentiment analysis, as well as visualization of the dependence of news sentiment on electricity imports and exports in Ukraine. The analysis of the results provides an understanding of the dependence of sentiment change compared to the previous months on electricity imports, as well as the trend of electricity imports in the face of constant negative news. The study concludes that the application of sentiment analysis together with visualization of data dependence in the energy sector is a valuable tool for determining the state of the energy sector and potential upward or downward trends. However, the sentiment analysis method used is expensive, and its application is relatively cheap on small amounts of data.