Publications

Abstract:

Finding the best approach to utilize customers' utility has always been discussed in the research. Initially, conjoint analysis was introduced in marketing as a survey-based way to learn more about consumers' desires. This technique was later supplemented by a discrete choice experiment (DCE) approach.  The DCE is a questionnaire where products are displayed to respondents as a set of alternatives. The respondents select the alternative that has the most utility. Multinomial or conditional logit models typically analyze the results of the survey. This analysis helps identify each attribute's impact on respondents' utility. However, the problem with the discrete choice experiment is its failure to consider the inherent uncertainty in customer responses. Regarding that problem, the discrete choice experiments are promoted to respondents express their preferences in a fuzzy number instead of choosing a specific option. This study focused on Iranian stock market traders expressing their preferences for purchasing company shares with fuzzy numbers. Additionally, a new model is developed by integrating standard discrete choice models with fuzzy sets to measure uncertainty in customer decisions. This approach allows us to understand the respondents' preferences and provides a more comprehensive model for decision-making in uncertain conditions.  

Abstract:

Insurance industries primarily rely on premiums as a source of income. Consequently, developing an accurate algorithm for predicting insurance charges is essential for these industries. The development of artificial intelligence and machine learning techniques has given rise to generating algorithms to predict insurance charges. This research presents a methodology to enhance the performance of the regression model, providing better outcomes than previous examinations. The performance of the regression model is analogous to the Random Forest and XG-boost algorithms. Although the regression method may not have better results in a mean absolute error, it performed better in the median absolute error value. Despite the simplicity of the regression models, they have demonstrated exemplary performance compared to previous studies and other models. 

Abstract:

Economic changes and recessions put pressure on citizens, and the buyers’ inclination to use second-hand commodities increases as a result. Cars are one of these goods, and trading second-hand cars is booming especially in developing countries, which is where buyers cannot afford a new car or people desire to make income by selling their assets. Determining the price may be challenging, so both buyers and sellers welcome a framework in regards to determining the price or identifying the influential features in order to determine the price. Several machine learning approaches, such as linear regression, ridge regression, lasso regression, random forest, and bagging are used in this study in order to provide a framework in regards to predicting the pricing of used cars. These techniques are regression-based, and their goal is to identify the regression line with the lowest absolute error value. Finally, the approaches that are used in order to provide buyers, sellers, and other participants in the second-hand automobile market with a solid statistical framework were carefully analyzed.