ANALYZING THE PAST AND MANAGING THE FUTURE USING NEURAL NETWORKS

An investigation is conducted on the possibility of analysing the underlying structure of a non-linear dynamical system by monitoring only time-series of past data with neural networks. From this analysis, it is shown that applying small perturbations to control variables can lead to an efficient control of state variables without any knowledge or a priori assumption regarding the system. Although this study is carried out with computer-generated data, the behaviour of the whole when random noise is applied to the system opens the possibility of a move efficient way of managing real-world complex systems such as those resulting from human activities.