Sharif Digital Repository

One of the important branches in machine learning is finding patterns and hidden features, which is useful for prediction and decision making in many cases. Finding patterns and features of financial data and particularly stock market data may play an important role in guidance, development and proper decision making related to such data sets. In this thesis, we focus on discovering and collecting appropriate methods in machine learning which can be helpful for gathering proper information from financial data and applying them on actual data  

In inefficient stock markets, fast and complete access to the public information about stocks and using this information for trades can make the investment more profitable. This research gathered the Iranian telegram channel's data with stock and investment subjects, trading volume, and stock returns. We suggested a trading strategy for beating the market by processing these data. We have also investigated the transaction costs in this research.
 

Since investing in the stock market is always known as one of the ways to increase capital, many researches have been done in the field of portfolio selection in order to provide methods to earn more profit and control investment risk. One of the influential factors in increasing the profit from the portfolio is the proper prediction of future of shares. Therefore, in many research, various methods and tools have been used to predict the future of shares more accurately. One of these tools is forecasting the market regime. In this research, harmonic patterns have been used to predict the market regime. Also, based on the harmonic patterns, the scope of entry into the transaction, the...