Forecasting plays a crucial role in decision-making and planning across various practical fields. To select the most suitable forecasting method, it is essential to examine the purpose of the forecasting, how the methods will be utilized, the determination of system components, and the impact of historical data on future predictions. Financial markets are evolving due to the growth and development of national economies, attracting the attention of many financial decision-makers for investment purposes. Despite the effectiveness of conventional methods, they all follow a distance-based approach in the learning process, which may not be sufficient to meet the objectives of decision-makers in financial markets. Another critical factor in financial market decision-making is the accuracy in correctly identifying the direction of price changes.This research introduces the construction of a neural network model for forecasting financial time series with the aim of achieving maximum possible profit. It demonstrates how to model a multilayer perceptron neural network based on a new direction-oriented approach. In the proposed model, two objectives are implemented in the model's objective function and its constraints: minimizing the distance error of predicted values and maximizing the number of directional matches in predictions. To eva‎luate the effectiveness of the proposed direction-based approach, five financial datasets were utilized, and the results obtained were compared with distance-based time series models.Experimental results indicate that the proposed approach outperforms traditional approaches in terms of profit measurement criteria such as net profit and return on investment, with average profitability percentages in the examined datasets being 0.16%, 0.25%, 0.05%, 0.22%, and 0.20% higher than traditional models. However, the results from measuring distance accuracy show that the error of the proposed model is greater compared to classic distance-based models. Therefore, the proposed model is a suitable alternative to classic price forecasting models in areas like the stock market, where profitability is prioritized over other forecasting objectives.

مهدي خاشعي آشياني

صبا صارمي نيا , حسين خسروشاهي