NEXUS: A multi-modal framework for capturing financial news interactions in market forecasting

Researchers are putting forth significant effort to develop accurate methods and techniques for stock market forecasting. It has emerged as a particularly difficult area of investigation. In addition to other cues, stock market prices are highly dependent on the mood of the stakeholders, which can be gauged through the analysis of news and social media data, as well as other indicators. A number of aspects of stock price prediction based on new data are discussed in this paper, which also attempts to establish a relationship between the polarity of news and the price of a stock for a given company. The experimental outcomes prove the capability of the proposed technique for predicting the stock prices for a certain point in time. The information used in the analytics has been gathered from seven different sources, including six financial news portals and one social media platform, according to the report. Additionally, various issues such as challenges, shortcomings, and future research directions have been discussed.

Big Data, as well as the visualization of data is a buzzword today. This publication surveys latest trends in visualization in financial industries. A very basic framework considering the different data, interaction, user, method, and evaluation methods is presented. Following a basic review of the existing literature in financial industry visualization, the framework is described in further depths, and shades insight into the different criteria of financial visualization in e.g. the contexts of regulation, market observation, prediction of market forecasting, and visualization of statistical information. The publication rounds up with a discussion of latest trends in media technology, and how these might impact the way how financial industries adopt these into their work-practices, such as virtual currencies, crowd-funding, social media trading, crowd-sourcing, or analysis of financial news. Thus, this publication manifested as a survey article about latest interactive media technology, and how this technology is enriching financial industries in many aspects for the purpose of visualization.

Financial markets are exhibiting more dynamism than ever before, and as a result, analytical frameworks and methodologies have been developed, not only to analyze but also to facilitate the navigation of volatility while providing a foundation for more informed investment decisions. This thesis describes a multi-data integration framework and application of computational methods for the purpose of stock market prediction to enhance the dependent nature of stock market decision making. This includes the application of Long Short-Term Memory (LSTM) neural networks to develop a means of comprehension of temporal dependencies of market data, and the application of Convolutional Neural Networks (CNNs) for technical indicator dependencies but with specific descriptions on how to ensemble and integrate LSTM and CNN to increase the predictive reliability of the models. Additionally, Natural Language Processing (NLP) methods are used to extract sentiment from headlines, articles, or social media references; and the multivariate time-series analysis includes macroeconomic indicator data time as well. Additionally, a reinforcement learning framework enhances portfolio methods, including a means of learning based on the temporal evolution of the market data in real-time. Aggregated to actionable conclusions to investors would require consideration of historical market validated pricing data obtained from market records, values of the variables from technical and fundamental aspects of analysis, and relevant literature involving money and liquidity. The aggregate of these results provided clear evidence of improved predictive performance of forecasting and disposition of probable market turning points and portfolio performance when compared to traditional analysis.