Bootstrapped Dendritic Neuron Model Artificial Neural Network for Forecasting

The dendritic neuron model artificial neural network, in which the dendrites in the biological neuron are added to the artificial neural network model, has been able to perform better compared to other artificial neural network methods used in the literature for the analysis of time series. The contribution of this study to the literature is the proposal of bootstrapped dendritic neuron model artificial neural network method as a new artificial neural network approach. The performance of the proposed method is compared with long-short term memory, a deep artificial neural network, pi-sigma, a high order artificial neural network and bootstrap hybrid artificial neural network according to various statistics on BIST100 time series. As a result of the applications, the proposed bootstrapped dendritic neuron model artificial neural network produced the best results in all statistics in general. In particular, the low standard deviation of the bootstrap dendritic neuron model stood out as a reason for preference compared to the other three methods. It is concluded that the bootstrap dendritic neuron model artificial neural network can be preferred to other artificial neural network methods for the analysis of time series.