• Citer

The public perception in designing and implementing Energy and Climate Policies. A machine learning sentiment analysis approach on Twitter

par Varelas, Panagiotis ;Jamshidiha, Mahdi ;Contino, Francesco;Parente, Alessandro
Référence ECOS 2023(25-30 June, 2023: Las Palmas de Gran Canaria, Spain), Proceedings of ECOS 2023, The 36th international conference on efficiency, cost, optimization, simulation and environmental impact of energy systems
Publication Publié, 2023-06-26

Publication dans des actes

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

Energy policies are an essential part of the energy transition, and public perception toward them is aquantitative indicator to evaluate them and plan them accordingly. Critical policies have recently been ratifiedaround the globe, from the Paris Agreement in 2015 to the European Green deal in 2019 and, more recently,the REPowerEU plan from European Commission following the War in Ukraine. Implementing such policiesaffects numerous aspects of the energy system, some of which could be quantitative, such as technology,electricity, and fuel costs. On the other hand, public perception is necessary, and we should discover how itcan affect energy policy and planning. This paper presents a quantitative analysis of twitter’s response to thelatest energy and climate policies. Our approach classifies tweets into two categories, one according to thetopic within three classes: Economy, Ecology, and Society, and the second based on the sentiment: Positive,Negative, and Neutral. Many algorithms are available in the literature for this study; we implemented kNN,SVC, Naïve Bayes, logistic regression, and random forest. Based on our findings, most tweets, around 80%,are neutral, with less being classified as positive and even less as negative. Regarding the subject category,classification is equally distributed within the three topics. The results could give policymakers insight into theirdecisions and their acceptance by the general population. We would extend our approach to unsupervisedlearning in the future to compare and improve algorithm performances.