The present volatility of the stock markets makes forecasting stock trends extremely challenging owing to several socio economic and political factors other than market trends. While machine learning models can be used to perform regression analysis based on historical numerical data trends, it becomes extremely challenging to incorporate the variabilities which are non-numeric in nature. Some of the factors which govern the rise and fall of stock prices are socio economic conditions, trade wars, current pandemic situation and global market slowdown, reliability of a company among others. Hence, one of the most effective ways to incorporate these trends is analyzing public trends pertaining to the same. While public sentiments may not always be coherent to prevailing market trends, yet they often portray the existential trends in the market and opinions of the public regarding potential purchases of stocks of a particular company in a given time period. This paper presents an approach which is an amalgamation of deep nets with attention and opinion mining for forecasting stock trends. The attention vector employed as an additional input computed on the moving average allows for current trend analysis along with opinion mining from public datasets encompassing both numeric data trends and non-numeric data parameters. The performance of the proposed system has been evaluated in terms of the error rates, regression and accuracy of forecasting for the system. Experimental analysis on benchmark S&P datasets show that the proposed approach outperforms baseline techniques in terms of accuracy of forecasting and regression.

Stock Market Forecasting, Deep Nets, Attention Vector, Opinion Mining, , Regression, Forecasting Accuracy