Public sentiment analysis of roadway work zones using social media data and machine learning models

The construction and maintenance of roadway infrastructure contribute positively to social and economic development and improve traffic safety. However, roadway work zones (WZs) present safety issues for construction workers and travelers, and adversely affect vehicular movement. By collecting and analyzing Twitter data (currently “X” threads), this study aims to explore public perceptions regarding WZs and identify factors that influence crashes and public experience at WZs. In this study, we employed several machine learning methods to classify WZ tweets and performed exploratory, sentiment, and emotional analyses of the classified tweets. We then verified our Twitter-related research outcomes using police crash reports. Sentiment and emotion analysis using classified tweets (with a 92% classification accuracy and 0.68 F1-score) showed somewhat negative emotions towards roadway WZs and onsite physical elements. However, the overall sentiment and emotion scores support the positive outcomes of WZ activities. We also found a strong temporal relationship between WZ-related tweets and fatalities. A cross-analysis of tweets and crash reports revealed that certain physical elements (e.g., signs, barriers, barrels, closures, and workers) are strongly associated with severe crashes at WZs. The results of this study may help policymakers to make appropriate policy decisions to improve driving experience and reduce WZ-related traffic accidents.