Intelligent structuration: Machine learning forecasting

Abstract

Artificial Intelligence and machine learning have contributed to today's business domain in areas including operational efficiency, smart decision making in cost savings, revenue maximization and customer satisfaction. However, there are missing links when machine learning algorithms are adopted in business environments. There is no rigorous and systematic explanation on data manipulation and machine learning algorithm selection processes. This research adopts the social construction perspective where the process of structuration is in the reciprocal interactions of data and algorithms using Structuration Theory. Intelligent Structuration for machine learning forecasting is proposed where three layers of structures - inference layer for signifying data, reference layer for dominating algorithms, and meta layer for legitimating insights - interact. The suggested intelligent structuration is demonstrated with hotel rate predictions. The empirical results indicate that the machine learning blending in the meta layer can significantly improve the prediction accuracy.