One of the most exciting tools that has entered our life in recent years is machine learning. The family of cognitive learning algorithms has already proved to bring convenience in many aspects considerably in our daily life. However, it is still unclear how machine learning can be utilized and create incremental value over existing information technologies and algorithms. This dissertation examines value creation with machine learning and compares its value to other competing systems. In the first two essays, we begin with an examination of the forecasting accuracy of machine learning algorithms compared to other traditional econometric models in predicting daily demand cases at a theme park. In the first essay (Chapter 2), we find improvements in machine learning algorithms' prediction accuracy in nonlinear and non-stationary daily demand forecasting. In the second essay (Chapter 3), we create a deep-learning neural network using a Human-in-the-Loop (HITL) approach. This chapter shows that the HITL approach outperforms prior approaches in prediction accuracy & consistency and in (in)direct economic benefits over other benchmarks. These results indicate that human expertise complements, rather than substitutes, machine learning. Also, the HITL approach performs better in demand prediction with sudden changes, bringing considerable advantages to business practices. In Chapter 4, we show the existence of price dispersion online. We leverage a machine learning application to measure information asymmetry in this research context. The analysis finds robust evidence of value creation with machine learning. The last chapter discusses implications and limitations for academia and practice for future machine learning applications, investments, and value creation research.