The presence of a substantial amount of customer feedback in the online space represents an extremely important area for businesses, as it provides insights and analysis that can contribute to improving their services. Automation of this process is crucial in handling a large volume of unstructured textual data. Therefore, sentiment analysis is one of the solutions for monitoring and analysing customer text comments. With this purpose in mind, our master's thesis describes the field of sentiment analysis, data warehousing for efficient data storage, and methods for data visualization. In an empirical study, we developed models for predicting the sentiment of customer comments in the area of gas station reviews, using various machine learning algorithms based on labelled data with appropriate text preprocessing. We placed the results of the best model into a data warehouse and created an automated report in the form of a dashboard. We found that text preprocessing improves the classifiers' results, and the support vector machine method yields the most effective predictions of sentiment in comments. We also presented the sentiment analysis data in an understandable manner, highlighting differences by year, country, company, and individual sales location.
|
