Agenda

Eshagh JAHANGIRI
PhD Student in Economics
36^th cycle

Full title: Stock Price Movement Prediction using Sentiment Analysis in Developing Countries: Evidence from Iran

Abstract:

One of the most important parts of financial decision-making is forecasting stock market movements. However, because stock market trends are dependent on so many internal and external factors—like political, economical, and environmental influences—predicting them is a challenging task. Moreover, stock price is determined not by investors’ own beliefs, but their beliefs about others’ beliefs. Two primary sources for capturing investors’ beliefs are web financial news and social media platforms. This paper explores how incorporating these sources for a stock movement prediction performs in a developing country and which of them has higher predictive power. A Support Vector Machine (SVM) model is developed to compare the performance of stock price movement prediction using web financial news and social media datasets from a developing country, Iran. The proposed model was tested on the Tehran Stock Exchange (TSE) market from January 1, 2020 to December 30, 2021. The findings show that for short-term predictions, web financial news sentiment has greater predictive power compared to social media sentiment. However, for medium- and long-term predictions, social media sentiment demonstrated better and more stable performance. In this paper, introduces, for the first time, two fine-tuned Large Language Models (LLMs) for the Persian financial context. The fine-tuned LLMs were trained on ParsBERT, a state-of-the-art pre-trained language model specifically designed for understanding and processing Persian (Farsi) text, based on the BERT architecture.

The seminar can be attended also remotely, connecting to ZOOM.

Link Zoom: bit.ly/insem-2425
ID riunione:  880 2639 9452
Passcode: InSem-2425