CHIA-YEN LEEHung Y.-HChen Y.-W.2022-04-262022-04-26202108848289https://www.scopus.com/inward/record.uri?eid=2-s2.0-85122424150&doi=10.1007%2f978-3-030-75162-3_4&partnerID=40&md5=3c7069f8ba07a5dbf55240595efd6a98https://scholars.lib.ntu.edu.tw/handle/123456789/607996This study proposes a hybrid data science (DS) framework and reinforcement learning (RL) in data envelopment analysis (DEA). The framework supports the functional form identification of the production frontier and the RL derives the optimal resource reallocation policy which guides the productivity improvement. In fact, both DS and RL techniques complement efficiency analysis. Emphasizes on planning over evaluation, we use data generating process (DGP) and an empirical dataset of power plants to drive productivity to validate the benefits of the hybrid DS framework and RL, respectively. Based on the results, we find that the hybrid DS framework and RL can enhance the interpretation of the production frontier and identify the optimal resource policy. ? 2021, The Author(s), under exclusive license to Springer Nature Switzerland AG.Data envelopment analysis (DEA)Data generating processData scienceReinforcement learningSymbolic regressionbook part10.1007/978-3-030-75162-3_42-s2.0-85122424150