Neural Networks for Time-Series prediction Application Field: Stock Exchange

スポンサーリンク

概要

• 論文の詳細を見る

• Time Series (TS) can be forecasted with a certain amount of accuratecy with Neural Networks (NN). In previous work a forecasting theory based on NNs was developed, for general TS as well as for the particular case of Stock Exchange (SE). SE Events were examined as Time-Series, solutions proposed for components: Trend, Cyclus, Season, Irregular events. This paper reports some implementation results of the above theory, as well as some theoretical improvements. For testing, a program with a Motif interface, has been developed. Data input has a three level hierarchy: (Level I:) stock market prices (buy/sell) for training, (Level II:) a previous weights matrix for testing of new, never processed, data and (Level III:) current stock market prices, for active forecasting (prediction). Some computing results are presented and comparison with former developed systems is performed. The main difference from previous systems consists in the mixt solution of mathematical and economical rules, that, the author hopes, will provide a better approach to the selected example field, as well as to others.

• 社団法人電子情報通信学会の論文
• 1997-03-06

著者

• Cristea A

  Univ. Electro‐communications Tokyo Jpn

• CRISTEA Alexandra

  電通大

• OKAMOTO Toshio

  電通大

関連論文

• Neural Networks for Time-Series prediction Application Field: Stock Exchange
• Stock Exchange Forecasting with the Help of Neural Networks

スポンサーリンク