Presenting a model for daily liquidity of stocks in Tehran Stock Exchange

Liquidity is a concept that is not clearly defined and so far more than 90 different criteria have been used for liquidity around the world. The present study aims to provide a local model for daily stock liquidity based on factors affecting liquidity. The values of 7 non-systematic factors that could be evaluated daily were extracted based on the data of 151 active companies in the period of 2009 to 2021 and divided into two clusters. Using structural equations with partial least squares approach, the validity of the identified variables was evaluated by the extraction criterion and their ability to explain its changes was calculated. Evaluation of the relationship between variables in machine learning models showed that final price, daily transaction value, daily returns, variance of daily returns and firm size have the greatest impact on clustering. Finally, the best model of machine learning was selected based on training and tests. The results show that the independent variables explain more than 83 per of the liquidity changes. Also, logistic regression model has a higher predictive power compared to other machine learning models and with 99.6 per fit accuracy, it is the most appropriate liquidity prediction model.