CRODNM: Chemical Reaction Optimization of Dendritic Neuron Models for Forecasting Net Asset Values of Mutual Funds

Learning algorithm and aggregation function used in the neurons have imperative influence on the approximation of an artificial neural network. Dendritic neuron model (DNM) using additive and multiplicative-based aggregation functions has been emerging as a machine learning approach and found successful in many engineering applications. This study attempts to advance the predictive accuracy of DNM through maintaining a decent steadiness between exploration and exploitation of its search space with chemical reaction optimization (CRO) algorithm, termed as CRODNM. The CRO, being a parameter free and powerful global search optimization method synergies with better approximation capability of DNM thus, able to overcome the limitations of conventional back propagation learning based DNM. In addition to this, to start the search operation with a better-quality initial population, we propose a new initial population generation method for CRODNM by incorporating several methods. The proposed CRODNM is evaluated on forecasting net asset values of four mutual funds in terms of convergence and prediction accuracy. The learning paradigm formed due to reasonable combination of CRO and DNM (i.e., CRODNM) found competitive and outperforms over DNM, multilayer perceptron (MLP), and genetic algorithm trained DNM prediction models.