A Comparative Study on Existing Techniques for Variable Reduction Including Factor Analysis, Principal Component Analysis, Correlation-Based Techniques, and Relief in Predicting the Risk of Stock Price Crash in Tehran Stock Exchange

In order to determine how well component extraction techniques (principal component analysis and factor analysis) and variable (feature) reduction techniques (correlation-based and relief techniques) perform in identifying the likelihood of future stock price crashes, the current study looked into these techniques' performance and effectiveness. To do this, a sample of 80 companies listed on Tehran Stock Exchange between 2006 and 2017 was chosen, and 17 often used major characteristics influencing the probability of a stock price drop were found by studying the literature. The criteria for evaluating the efficacy of the procedures under consideration included the mean absolute magnitude percentage error, root mean square error, and coefficient of determination. In comparison to the use of all key explanatory variables, the results showed that variable reduction and component extraction strategies work significantly better and are more effective at predicting the likelihood of future stock price crashes.