نوع مقاله : پژوهشی - کاربردی
نویسندگان
1 گروه آموزشی طراحی شهری، دانشکده معماری و شهرسازی، دانشگاه هنر ایران، تهران، ایران
2 گروه طراحی شهری، دانشگاه هنر ایران
چکیده
کلیدواژهها
عنوان مقاله [English]
نویسندگان [English]
Among the proposed methods for sentiment analysis methods based on social virtual space data and data mining are relatively new methods that have been used in this research. Google Map and textual data has been used as a source to extract users' opinions from Mellat Park (Tehran). The purpose of this article is to compare the methods of urban sentiment analysis in Mellat Park and to present a predictive model of users' sentiments from urban green spaces in Tehran.
This research is analytical and based on a quantitative method (supervised machine learning and lexical-based). The number of comments used on the Google Map social network after balancing with Smote technique was 1692 records. After pre-processing and labeling the data sentiments have been examined and analyzed with two methods (model-oriented and non model-oriented) with Python programming language.
The comparison of these two methods showed that among the machine learning algorithms, XG Boost with the highest accuracy (87 percent), K-Nearest Neighbors and Support Vector Machine with less accuracy are capable of predicting sentiment in green spaces. Compared to machine learning, the non-model-oriented (which uses VADER as a rule-based method) is less predictive. Finally, by using a Blending Learning model of the Stacking type (which combines several weak learning algorithms with Meta learning) has been used, which according to the results of the confusion matrix, it has the ability to predict better with 96% accuracy.
کلیدواژهها [English]
)