Document Type : Original Research Paper
Authors
1 Department of Statistics, Faculty of mathematical sciences, Vali-e-Asr University of Rafsanjan, Rafsanjan, Iran
2 Technical Deputy of Non-Life Insurance, Saman Insurance, Tehran,
Abstract
BACKGROUND AND OBJECTIVES: One of the criteria for deciding to invest in a listed company is the amount or changes in the stock price of the company in the future days and months. Various methods have been studied to predict the stock price and investment risk in a company. In most of these methods, the stock price is predicted as a continuous response variable. For this purpose, time series models are used in which assumptions such as the normality of disturbances or the linearity of the model are important. The purpose of this research is to introduce a two-category response variable based on the direction of share price movement in the next day and to introduce some statistical classification methods to predict it. These models do not have the limitations of the previous models, and for that reason they are of interest. The main objective of this article is to implement the studied methods and compare their accuracy in predicting the orientation of stock price movement of stock exchange insurance companies.
Methodology: In the current research, we have predicted the direction of stock price movement by using K-nearest neighbors, decision tree and random forest algorithms, which are among the non-parametric classification methods of statistical learning. The data used in this research includes information on the stock price of one of the insurance companies during the years 2019 to 2020, which has a suitable and high share in the portfolio of the insurance industry. To determine the accuracy of the studied models, the data were randomly divided into two groups, training and testing. Then, the statistical learning models were implemented on training data and their validity was measured using experimental data.
FINDINGS: The research results indicate the high accuracy of all three non-parametric models in predicting the stock price category of the insurance company. Likewise, among the studied models, the K-nearest neighbors algorithm performed better than other algorithms in predicting the direction of stock price movement.
CONCLUSION: Considering the importance of the risk of investing in an insurance company for customers, attainment to a valid model for stock price classification and specifying the variables that increase or decrease the price can help customers and insurance companies make better decisions.
Keywords
Main Subjects
Letters to Editor
Send comment about this article