Spatial federated learning approach for the sentiment analysis of stock news stored on blockchain

Abstract

Sentiment analysis can be a useful tool in predicting stock market trends, as it allows us to gauge the overall sentiment towards a particular stock or company. When combined with news stored in a blockchain, sentiment analysis can provide a more accurate and trustworthy representation of the market. The aim of this paper is to study the spatial news and external events which disrupt the stock market movement as well as news analytics techniques to understand the impact of news by spatial sentiments on stock market movement. For the prediction of stock market trading decisions, a novel ensemble technique consisting of spatial federation of deep learning algorithms, machine learning algorithms and dictionary based approach is proposed to maintain privacy and geographic diversity. These learning techniques are federated into five groups and these groups are ranked as per the prediction accuracy of these models. A bit 1 or 0 is assigned for each federation thus creating a 5 bit pattern which can be used to predict stock market trading decision as strong buy, buy, hold, strong sell, sell.