A STOCK PRICE PREDICTION MODEL BASED ON INVESTOR SENTIMENT AND OPTIMIZED DEEP LEARNING

Abstract

The MS-SSA-LSTM model is introduced in the study; It improves stock prices by combining emotional analysis, deep learning techniques, flock -tinting algorithms and data from many sources. To create a kind of emotional dictionary and calculate an emotional index, this model uses emotional analysis on the posts made for the East Money Forum. This provides useful information on how the market spirit affects stock prices. To improve the accuracy of predictions, the Sparrow Search Algorithm (SSA) is used to accommodate the LSTM hypermeters. Experimental findings clearly show the extraordinary performance of the MSSA-LSTM model. This is a great resource for creating reliable stock prices. The model is well suited to China's unexpected financial market and provides useful information for investors' dynamic decisions in the nearest period as a share price estimates. In addition, a model that combines LSTM and GRU was presented with the aim of classifying the warehouse mood. In addition, a strong outfit approach was used, including a voting eligible for emotional analysis and a voting -eligible retrograde to predict the share value. Using these dresses, which were easily merged with already existing models (MLP, CNN, LSTM, MS-LSSA-LSS), the total prognosis improved 0.999, or 99.9%. A user -friendly flask framework was designed with SQLITE support to facilitate the user's involvement and test signs, log on and model assessment procedures.