Considering for the Forecasting Method of the Nikkei Stock Average using Deep Learning with Natural Language Processing  
Author Ryota Nekomoto

 
Co-Author(s) Tomoaki Akiba

 
Abstract This study aims to improve the prediction accuracy of the Nikkei Stock Average thereby facilitating the determination of financial products to optimize asset building. We considered a prediction method for the Nikkei Stock Average using Long Short Term Memory (LSTM) and Convolution Neural Network (CNN) among deep learning techniques. The LSTM learns the open, high, low, and close (OHLC) values for x days and predicts the closing price for x +1 days. In the CNN, one value on day x +1 is added to the data trained in the LSTM and trained again. The Nikkei Stock Average data were obtained between 1988 to 2024 using yfinance, an open-source tool that uses Yahoo's public API. As a result, all LSTM models, except the one using Adaptive Moment Estimation with Weight decay (AdamW), achieved accuracy rates higher than the chance rate. The LSTM models using all optimization algorithms, except Stochastic Gradient Descent (SGD) and AdamW, showed the highest accuracy. In the CNN model, the prediction accuracy decreased for all algorithms except AdamW when compared with the LSTM model. These results indicate that it is possible to make predictions that can be used to assess economic conditions to a certain extent.

 
Keywords Forecasting Nikkei Stock Average, Data Collection and Analysis, Data Mining, Deep Learning, Algorithms, Optimization
   
    Article #:  RQD2025-231
 

Proceedings of 30th ISSAT International Conference on Reliability & Quality in Design
August 6-8, 2025