Document Type : Article
Authors
1
رییس دانشکده فنی مهندسی دانشگاه میبد ، گروه مهندسی صنایع (سیستم ها) ، دانشگاه میبد ، میبد ، ایران
2
Industrial Engineering Department, Meybod University
3
Assistant Professor, Department of Industrial Engineering, Faculty of Engineering, Meybod University, Meybod, Iran
10.24200/j65.2024.63158.2375
Abstract
Increasing profitability and reducing risk always requires choosing a smart investment path while taking advantage of data analysis; Therefore, it is necessary to provide a technique in the form of a decision support system for stock portfolio management while better understanding the position of data science. In this research, while combining data science methods with the Markowitz model, classification and forecasting models have also been created. Detecting financial fraud of companies is also effective; Hierarchical and Cummins algorithms have also been used in order to cluster active companies in Tehran Stock Exchange. In terms of data classification, the linear support vector machine classification algorithm has been identified with 70% accuracy in comparison with the decision tree, Nyobies, nearest neighbour, and multilayer perceptron algorithms, and in terms of building prediction models, the decision tree with the minimum amount of error, in comparison with Nyobies algorithms, is the closest Neighbor, multilayer perceptron has been identified. The primary data in the current research includes twenty titles of financial indicators and daily data of stock prices of companies in the Python programming space. It was concluded that choosing stocks from among the clusters formed by different stocks of companies will reduce the risk of diversifying the stock portfolio. The mentioned system will be able to generalize as a comprehensive model for data in different time intervals and the indicators desired by the analyst, thus helping investors in a fast and accurate analytical way. The performance of the final technique will be such that the investor first selects a number of desired companies according to the heard, analysis and news; It predicts the price performance of companies; Then, it separates the companies that have succeeded in accepting the condition by using the clustering model and separates the portfolio of various stocks. It forms according to the amount of risk and expected return. Importantly, as mentioned, the presented classification and forecasting models, in addition to being used in the formation and management of the stock portfolio, will be able to be effective in predicting the possible fraud of companies.
Keywords
Main Subjects
)