Role of Prediction in Data-driven Pricing Schemes

Data creates true values to the society, which urges the application of data-driven approaches in every aspect of our lives. While data-driven approach is indeed promising, much remains unknown. Take electricity market design as an example, the success of data-driven pricing schemes may heavily rely on the accurate prediction of market conditions. In this paper, we explore the role of prediction in data-driven pricing scheme design. Specifically, we employ two prediction models: Autoregressive Integrated Moving Average (ARIMA) model and Long Short-Term Memory (LSTM) model. We use numerical studies to highlight how the prediction error could affect the vulnerability and robustness of the pricing scheme. We submit that a good predictor could contain the vulnerability of the data-driven pricing schemes while an ill-designed predictor may amplify the market loopholes.