MCMC Approach for Stock Price Forecasting Using an Italian-BERT Model

Sentiment Analysis is a task of Natural Language Processing (NLP) whose main goal is to classify sentences (or entire texts) to obtain a score about their polarity: positive, negative, or neutral. Recently, a Transformer-based architecture, AlBERTino [5], has been introduced to determine a sentiment score in the financial sector through a specialized corpus of sentences. Here, the AlBERTino model can be used to improve stock forecasting, determining the sentiment score associated with events in the market and using a Markov Chain Monte Carlo (MCMC) method to determine a new series of bounded drift and volatility values based on this score. With these new values obtained through Bayesian inference, generating a series of paths through a Monte Carlo method to predict a polarity-driven future price is possible.