نوع مقاله : مقاله علمی پژوهشی
نویسندگان
1 کارشناس ارشد مهندسی مالی، دانشگاه میبد، یزد، ایران
2 استادیار، گروه مهندسی صنایع، دانشگاه میبد، یزد، ایران
3 دانشیار، گروه مهندسی کامپیوتر، دانشگاه میبد، یزد، ایران
چکیده
کلیدواژه‌ها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Objective: Knowing the behavior of the capital market and their deviations is a background for analyzing the behavior of returns during the occurrence of events governing the society. The political, economic and social issues of the world can easily affect the economic cycle. The stock market, as an important part of the economy, will not be exempt from this. The high accuracy of predicting and recognizing fluctuations will increase investor confidence and lead to correct and timely decisions for asset management. Also knowing the most efficient tool for predicting returns is necessary to analyze the behavior of this market. The purpose of the current research is to cluster the companies in the stock market according to their impact on the events of the sanctions period using the best artificial intelligence method for forecasting.
Methods: The weekly return data of 200 active companies in the Iranian stock market, information related to the variables of industry type, size, liquidity and profitability of the selected companies in the period from 2016 to 2021 along with selected political, economic and social events have been used in this research. In the first step, four models LSTM (Long-Short Term Memory), DQN (Deep Q Network), RF (Random Forest) and SVR (Support vector machines) were compared as the best models of deep learning and machine learning And then the stock returns are predicted based on the superior model. In the second step, the sensitivity analysis of the scenarios resulting from the effectiveness of the changes in the efficiency with respect to each of the inputs of the industry type, size, liquidity and profitability of the companies are carried out and finally, the clustering of the results in three categories of economic, political-economic and economic-social events has been done by using the separation clustering method.
Results: LSTM (Long-Short Term Memory) is a superior model to other deep learning models (LSTM,DQN) and machine learning models (SVR,RF) for predicting stock returns. The results of a wide range of analyzes are available to investors depending on their needs, which can be used as a basis for the analysis of the return process when faced with events. But in general, it can be said that political events have the greatest impact on the returns of companies' shares. After that, economic events and finally social events have the least impact on the returns of companies' shares. In order to evaluate the criteria, company size, type of industry, liquidity and finally profitability have taken the last place in the ranking of factors affecting fluctuations.
Conclusion: Iran's stock market is influenced by political, economic and social news as well as government actions and statements, but depending on the type of news, their effectiveness will be different.The impact of events on the returns of companies' shares is direct and the truth of this statement that during the occurrence of events, the returns of stock companies fluctuate depending on the type of industry, size, liquidity and profitability of the companies is confirmed. Meanwhile, political events have the greatest impact on the returns of companies' shares and should be paid attention to by capital market activists.
کلیدواژه‌ها [English]
)