This paper presents a novel technique to forecast the price trend (direction) of 25 different commodities, listed on international markets, using a neuro-fuzzy controller. The forecasting system is based on two independent adaptive neural fuzzy inference systems (ANFISs) that form an inverse controller for each commodity. The ANFIS controller belongs to direct control and is based on inverse learning, also known as general learning. Daily data return sets, for the period 14th October 2009 until 28th September 2012 for 25 different commodities, are used to learn and evaluate the proposed system. The results of the trading simulation and the experimental investigations carried out in the laboratory are provided. The forecast accuracy of the proposed technique is evaluated by out-of-sample tests. The return on equity based on the hit rate and the comparison of equity with the buy and hold strategy are the central evaluation criteria. The results are very encouraging, showing high accuracy of the hit rate reaching 68.33 % and a notable superiority of the return on equity when compared with the buy and hold strategy. Also the performance of the model is compared with that of other approaches.

Since accurate forecasting of energy export is very important for planning potential energy demand and improving the energy production sector, various forecasting methods have been developed. The present work is focused to apply a novel technique, an integrated neuro-fuzzy controller named PATSOS. The forecasting system is based on two Adaptive Neural Fuzzy Inference Systems (ANFIS) that form an inverse controller. An ANFIS model represents the controller and another ANFIS represents the energy export model that is going to be controlled. ANFIS uses a combination of the least-squares method and the backpropagation gradient descent method to estimate the optimal energy export forecast parameters. The ANFIS controller belongs to direct control and is based on inverse learning, also known as general learning. Hourly data sets during the period were used to learn and evaluate the proposed system. The forecast accuracy of the proposed technique was evaluated using out of sample tests. The results of the simulation based on statistical errors and the experimental investigations carried out on the laboratory showed that the model despite the high data volatility, is suitable for forecasting hourly energy exports.