Measurement and quantification of an individual's feelings for a place in personal data analysis

Location-related data are an important subset of personal data. An individual may have a positive or negative feeling for a specific place, which is important for personal data analysis. There are many studies on sentiment analysis within text data, such as tweets, but few studies have been conducted specifically on an individual's feelings regarding locations. In this study, we focus on measuring and quantifying an individual's feelings for a place using three representative methods in sentiment analysis: emotion dictionary, personalized dictionary, and Bayesian classification. We design an experiment to evaluate these methods using tweet data including locations and an individual's emotional changes with regard to these locations before entering, after exiting, and in a location. Three sets of emotion scores are obtained and normalized. Furthermore, we set four protocols and use statistical methods to compare these emotion scores with the subjective emotion scores provided by the user whose tweets are used in the experiment. Experimental results show that Bayesian classification performs the best in measuring and quantifying an individual's feelings for a place.