Showing 6 results for Machine Learning
Yasin Ghasemi, Abbas Khandan, Narges Akbarpour-Roshan,
Volume 13, Issue 47 (5-2022)
Abstract
The pension coverage of the Iranian Social Security Organization for self-employed workers is offered at three contribution rates of 12, 14 and 18 percent, but looking at the statistics shows that the demand for these types of insurances is low. This research investigates the characteristics of these insured groups by using data mining and applying two machine learning algorithms, decision tree and random forest, and predicts their behavior by providing a classification model. This will help the Social Security Organization to improve customer relationship management. For this purpose, the information of 1286174 insured persons of self-employed in 2020 was used, which includes the characteristics of age, gender, average monthly income, the years of service, and the type of self-employed pension scheme. The obtained results show that women mainly apply for the scheme with 12 percent contribution, while men tend to be covered by schemes with contribution rates of 14 and 18 percent due to the burden of supporting the family. Also, for men, the demand for schemes of 14 and 18 percent increases with the increase of age, income and years of service, but there are no such trends for women. According to the obtained results, years of service and then gender are decisive in choosing the type of pension scheme in such a way that according to the prediction of the model, people with less than 4.5 years of service are known as definite applicants for 12 percent self-employed pension scheme.
, Abbas Khandan,
Volume 14, Issue 52 (9-2023)
Abstract
Purpose: The aim of this study is to identify and classify insurance customers in order to identify the target population for increasing the profitability of insurance companies, achieving a balance in premium payments, and examining the health questionnaire as an indicator of policyholders' preferences. Moreover, designing a marketing strategy to optimize advertising efficiency.
Method: In this paper, five machine learning algorithms, namely Decision Tree, Random Forest, Support Vector Machine, Naive Bayes, and Logistic Regression, are used to classify customers into two categories: profit-generating and loss-generating. Data from a private insurance company is utilized, consisting of 2,897 observations collected from December 1400 to December 1401.
Findings: By utilizing machine learning methods and focusing on the target population, the chances of success can be increased. The presence of a small number of individuals who significantly reduce the profitability of insurance companies is evident. The pre-existing medical conditions of individuals have a considerable impact on their insurance usage and the damage caused to insurance companies.
Conclusion: Machine-learning methods can provide a comprehensive understanding of insurance customers and their needs. By identifying the target population, insurance companies can increase their profitability and satisfy their customers by addressing their specific demands
Fatemeh Ansari, Shahab Jahangiri, Ali Rezazade,
Volume 14, Issue 53 (12-2023)
Abstract
Objective: The aim of this research is to provide a practical guide for investing in the Tehran Stock Exchange by combining technical analysis techniques with advanced machine learning methods. Focusing on the analysis of buy and sell signals in selected indices of the Tehran Stock Exchange, the study seeks to evaluate the effectiveness of machine learning models in predicting market trends.
Materials and Methods: In this study, the daily data of six selected indices of the Tehran Stock Exchange, including financial, petroleum products, automotive, pharmaceutical, food, and basic metals indices, were analyzed from 2020 to January 2025. Four machine learning models, including Linear Model, Random Forest, Artificial Neural Network, and Support Vector Regression, were utilized alongside two technical analysis strategies, TEMA and MACD, to generate and evaluate buy and sell signals.
Results: The results indicated that machine learning models, particularly Random Forest and Artificial Neural Network, performed better in identifying buy and sell signals when combined with TEMA and MACD strategies. These models were able to predict market trends with higher accuracy, and the signals they generated were mostly consistent with actual price changes. The food, automotivation and financial and basic metals indices demonstrated greater sensitivity to these analyses.
Conclusion: The combination of machine learning methods with technical analysis strategies can provide investors with a powerful tool for decision-making in the Tehran Stock Exchange. This research demonstrated that using these methods can not only improve the accuracy of buy and sell signals but also reduce investment risk and increase returns. Utilizing these models can be recommended as part of an investment strategy for analysts and investors.
Originality: This research is the first quantitative study that seeks to conceptualize buy and sell signals using the combined method of machine learning and technical analysis as one of the basic tools to guide investors. |
Majid Shafiei, Parviz Rostamzadeh, Mohammad Rastegar, Zahra Dehghan Shabani,
Volume 14, Issue 53 (12-2023)
Abstract
The stock market, as one of the vital components of the capital market, is an important part of the country's economy that can manage the flow of capital, optimize capital allocation, and thereby contribute to economic growth and development. More accurate prediction of the stock market trend can help investors' decision-making for higher returns by reducing risk. In general, the stock market is constantly changing and many factors influence the trend of this market, so predicting the patterns of movement in the stock exchange requires sufficient information about the past and influencing factors of the market. This article is part of the forecast of the stock market index of Iran, seeking to interpret the model and identify the most influential economic variable on the price index prediction. For this purpose, daily stock market and economic data, during the period 1394-1401 were used. Machine learning models are also used for prediction and the Shapley Additive exPlanations (SHAP) to interpret how to predict and determine the most important variables in the predictive model. Based on results from tree-based ensemble methods, the proposed model in this study, ExtraTrees, performed best based on predictive error criteria. In the study of the feature importance is also based on the ExtraTrees model, in order of the dollar rate (Nima), unemployment rate, dollar rate of market and liquidity, the most important economic variables influencing the forecast model. Also, according to other models used in the research, liquidity is the most effective variable on the stock index trend. Finally, it can be said that the most effective monetary variables on the stock market index in Iran are liquidity and exchange rate variables, so monetary policymakers and stock market investors should be more sensitive to these variables in their decisions.
Ali Moridian, Hassan Heidari, Seyed Mehdi Hosseini, Heshmatollah Asgari,
Volume 16, Issue 60 (9-2026)
Abstract
Objective: This study examines the effects of economic policy uncertainty, exchange rate, and oil price on inflation in Iran during the period 2008 to 2023. The main objective is to identify the short-term, medium-term, and long-term nature of these effects and analyze inflation dynamics using modern wavelet and machine learning methods.
Materials and Methods: Regularized least squares regression with wavelet kernel (WKRLS) and nonparametric wavelet quantile causality (WNQC) are used to analyze nonlinear and scale-dependent relationships between variables. The data include inflation index, economic policy uncertainty (EPU), unofficial exchange rate, and oil price on a monthly basis. The generalized wavelet quantile Dickey-Fuller test (Wavelet-QADF) is also used to examine the stationarity of time series.
Results: The results show that key variables of the Iranian economy are stationary in most quantiles and time scales. According to WKRLS estimates, the effect of economic policy uncertainty on inflation is weak in the short run, decreasing but still significant in the medium run, and increasing non-linearly and acceleratingly in the long run. The exchange rate has the greatest impact on inflation, especially in the short run due to the Iranian economy’s heavy dependence on imports. Oil prices also have a significant impact on inflation and its volatility in the long run. WNQC findings show that economic policy uncertainty and exchange rate uncertainty have a stronger effect in the low and middle quantiles of inflation, while oil prices mainly amplify inflation fluctuations in the long run.
Conclusion: The findings emphasize the importance of stable economic policies, reducing dependence on oil revenues, and controlling exchange rate fluctuations for managing inflation in Iran. Also, combining wavelet and machine learning methods allows for a more comprehensive analysis of inflation dynamics in different conditions.
Mahmood Mahmoodzadeh, Masood Soufi, Morteza Alipour,
Volume 16, Issue 60 (9-2026)
Abstract
Despite the growing use of machine learning in credit scoring, many domestic studies still rely mainly on traditional statistical models and static borrower characteristics, while limited attention has been paid to the role of real behavioral, transactional, and repayment-performance data in post-disbursement credit risk monitoring. To address this research gap, this study compares the performance of four stepwise logistic regression models and the LightGBM algorithm in predicting credit default, using data from 119,050 loan facilities granted to individual customers of Resalat Qard al-Hasan Bank during the period between 26 March 2022 and 18 March 20240. The target variable was defined based on repayment delays of more than 90 days, and model performance was evaluated using AUC, Accuracy, Recall, F1-Score, and Balanced Accuracy. The knowledge contribution of this study lies in providing empirical evidence on the effectiveness of real banking data, focusing on behavioral and transactional variables, comparing a classical statistical model with a machine learning algorithm, and assessing model performance in identifying the minority class under imbalanced credit data. The results indicate that repayment-related variables, particularly the number of overdue installments and outstanding debt balance, are the most important predictors of default. Although the fourth logistic regression model achieved a high overall AUC of 0.98, it performed poorly in identifying high-risk customers, with a Recall of only 0.12%. In contrast, LightGBM identified 92.2% of high-risk customers and outperformed logistic regression on imbalance-sensitive evaluation metrics. These findings suggest that, in imbalanced credit datasets, relying solely on AUC and Accuracy can be misleading, while Recall, F1-Score, and Balanced Accuracy are more informative for assessing a model’s ability to detect high-risk borrowers. Therefore, in the post-disbursement monitoring scenario, machine learning algorithms based on behavioral and transactional data can provide a more accurate and reliable framework for credit risk management in Iranian banks.
