Using Neural Networks for Generic Strategic Planning

Abstract

This paper argues the scarcity of strategic planning software is due to Western philosophical traditions which see strategy hypothesising as a mysterious, intuitive process that resists analysis and computerisation. But progress is possible if one extracts, from the tactical planning literature, eight key, generic, strategy-evaluation criteria. An experiment then tests whether scores on such criteria can be used to power machine learning of overall strategy desirabilies, and whether such learning is better achieved using multiple regression analysis or a simulated neural network. Both methods were successful, but the neural network was clearly the most accurate. It therefore constitutes a promising basis for self-improving, strategic planning software.