|
|
|
 |
Search published articles |
 |
|
Showing 3 results for Refahi Sheikhani
Dr. Mehrdad Fadaei Pellehshahi, Dr. Sohrab Kordrostami, Dr. Amir Hosein Refahi Sheikhani, Dr. Marzieh Faridi Masouleh, Dr. Soheil Shokri,
Volume 11, Issue 2 (2-2020)
Abstract
In this study, an alternative method is proposed based on recursive deep learning with limited steps and prepossessing, in which the data is divided into A unit classes in order to change a long short term memory and solve the existing challenges. The goal is to obtain predictive results that are closer to real world in COVID-19 patients. To achieve this goal, four existing challenges including the heterogeneous data, the imbalanced data distribution in predicted classes, the low allocation rate of data to a class and the existence of many features in a process have been resolved. The proposed method is simulated using the real data of COVID-19 patients hospitalized in treatment centers of Tehran treatment management affiliated to the Social Security Organization of Iran in 2020, which has led to recovery or death. The obtained results are compared against three valid advanced methods, and are showed that the amount of memory resources usage and CPU usage time are slightly increased compared to similar methods and the accuracy is increased by an average of 12%.
Dr. Mehrdad Fadaei Pellehshahi, Prof. Sohrab Kordrostami, Dr. Amir Hossein Refahi Sheikhani, Dr. Marzieh Faridi Masouleh, Dr Soheil Shokri,
Volume 13, Issue 2 (12-2022)
Abstract
In this paper, a new method is presented using a combination of deep learning method, specifically recursive neural network, and Markov chain. The aim is to obtain more realistic results with lower cost in predicting COVID-19 patients. For this purpose, the BestFirst algorithm is used for the search section, and the Cfssubseteval algorithm is implemented for evaluating the features in the data preprocessing section. The proposed method is simulated using the real data of COVID-19 patients who were hospitalized in treatment centers of Tehran treatment management affiliated to the Social Security Organization of Iran in 2020. The obtained results were compared with three valid advanced methods. The results showed that the proposed method significantly reduces the amount of memory resource usage and CPU usage time compared to similar methods, and at the same time, the accuracy also increases significantly.
Amirhossein Malakouti Semnani, Sohrab Kordrostami, Amirhossein Refahi Sheikhani, Mohammad Hossein Moattar,
Volume 16, Issue 2 (8-2025)
Abstract
Insurance companies face the critical challenge of identifying “good customers”—policyholders who consistently pay premiums with minimal or no claims—within large, heterogeneous datasets. This study proposes and evaluates a hybrid machine learning framework to predict good customer status using an enhanced insurance dataset that integrates demographic, financial, and policy-related features. The framework combines an XGBoost classifier, a soft-voting ensemble of RandomForest and LightGBM, and a custom Transformer Encoder, with all models tuned using the Optuna hyperparameter optimization library to enhance predictive accuracy and interpretability.
The methodology includes preprocessing steps such as categorical encoding and standardization of numerical variables (e.g., age, BMI, premium with GST), followed by a novel label engineering scheme that defines good customers as those whose premiums exceed the mean plus one standard deviation and have no claim history. The dataset is split into training (80%) and testing (20%) subsets. Two hybrid architectures are developed: Model A, which fuses the predicted probabilities from XGBoost and the Transformer Encoder using a 60–40 weighting, and Model B, which employs a soft-voting ensemble of RandomForest and LightGBM. Ablation studies quantify the contribution of each component, while performance is assessed using accuracy, AUC, F1-score, and Matthews Correlation Coefficient, supported by visual tools such as correlation heatmaps, ROC curves, and confusion matrices.
Experimental results show that Model A attains an accuracy of 0.8720 and an AUC of 0.9140, whereas Model B achieves an accuracy of 0.8850 and an AUC of 0.9260 after systematic hyperparameter tuning. Removing either the Transformer or XGBoost markedly degrades Model A, while omitting RandomForest or LightGBM leads to smaller performance drops in Model B, underscoring the value of ensemble diversity. Overall, the proposed framework provides a practical tool for insurance customer segmentation and profitability-oriented decision-making, and its open-source implementation facilitates replication, extension with additional features or larger datasets, and potential real-time deployment in operational insurance environments.
|
|
