APPLICATION OF NEURAL NETWORKS FOR PREDICTING THE VALUE OF SECURITIES

Solobuto Aleksei Viktorovich, Pavlov Valeriy Anatolyevich

This paper explores the application of neural networks as a tool for predicting the value of securities based on historical data. To train the model, data preprocessing is carried out, including the construction of time series, which allows for the identification of hidden dependencies between variables. To improve the model’s robustness to outliers and reduce the risk of overfitting, regularization methods such as Dropout and L2 normalization are employed. In addition, appropriate activation functions and a loss function were selected to address the specific task. The study demonstrates the potential of neural network approaches in financial forecasting and highlights the importance of selecting the right architecture and tuning model parameters according to the characteristics of the input data.

Keywords:neural networks, time series, regularization, stocks, indicators.