Development of geothermal power plants and local geothermal energy initiatives have lagged due to the social problems such as conflicts with stakeholders such as Onsen (Hot Spa) owners, despite the abundant geothermal resources. Study area was Tsuchiyu Onsen in Fukushima prefecture, Tohoku (Northeast) District, Japan, where the Great East Japan Earthquake and Fukushima Nuclear Accident occurred in 2011, and the reconstruction and local initiatives of geothermal energy were still unclear. Agent-based modeling (ABM) is an effective methodology for modeling and analysis of opinion formation. Parameter estimation method was proposed to extract appropriate parameters from various factors through a Bayesian Network. The characteristics of stakeholders and communities that affected opinion formation in the survey area were successfully extracted. Here we show the sufficient methodology to quantify the characteristics of each person using survey data, and to extract the parameters of the agent by data-driven inverse analysis. By using this methodology, we could reproduce opinion diversity, which is a property of opinion formation in real communities. This result suggests that the model replicates the actual formation of opinion in Tsuchiyu, where the economy was boosted by the construction of a binary cycle power plant.