Iranian Journal of Operations Research

A Hybrid Bayesian BWM–Fuzzy MARCOS–Metaheuristic Framework for Sustainable Smart Locker Location in Last-Mile Urban Logistics

Abdollah Arasteh

The explosive growth of global e-commerce and the increasing complexity of last-mile logistics have made the strategic placement of smart lockers a critical concern in modern urban logistics systems. Conventional methods, which rely solely on Multi-Criteria Decision Making (MCDM) methods for obtaining solutions, suffer from several limitations when implemented in uncertain, significant, and multi-objective scenarios. This paper proposes a stochastic multi-objective optimisation model for the BWM, prioritising decision criteria, which is solved by combining a hybrid metaheuristic solution methodology. The proposed model optimizes both total cost and sustainability performance from economic, environmental, and social perspectives, as well as robustness to demand uncertainty. An empirical study using Babol City, Iran, is presented to test and demonstrate the proposed framework. Candidate locker location and demand areas were examined based on expert-elicited criteria weights, with the preparation of a multi-objective mixed-integer programming model. In order to alleviate the computation burden, a combined structure of NSGA-II and LNS (referred to as NSGA-II+LNS) was proposed, which outperforms classical evolutionary algorithms in terms of convergence into the Pareto frontier. Factual results indicate that factoring in economic affordability, accessibility, and environmental impact is key to optimal locker capacity design. Robust solutions under demand fluctuation can save up to 18% more on service reliability, providing strong deterministic answers. This article makes the following theoretical and practical contributions: (i) a novel sustainable-oriented, deterministic model for smart locker location is proposed; (ii) advanced metaheuristics are integrated with MCDM in urban logistics, whereas fewer studies have focused on integrating them; and (iii) policy implications are suggested not only to policymakers but also to logistics operators who want robust last-mile delivery strategies..

Classification of Leukemia Using a Hybrid Approach Based on Temporal Fusion Transformer and XG-Boost

Elham Askari

Leukemia is one of the most common and dangerous types of cancer in the world. In many cases, the disease is curable if detected in its early stages. One of the effective tools for early detection is the analysis of microarray data, which measures the expression of thousands of genes simultaneously. However, the large volume of features and the presence of noise make the analysis process complex and time-consuming. Therefore, the selection of effective genes plays a key role in increasing the accuracy and reducing the computational cost of learning models. In this paper, a two-step hybrid approach is presented for feature selection and classification of leukemia types. In the first step, the features are filtered using the mutual information criterion and the genes with the highest correlation with the disease label are selected. In the second step, the XGBoost model is used to rank and stably select the features to identify the genes that are most important in different iterations. In the final stage, classification will be performed using the temporal fusion transformer method, which allows for fast and efficient learning of complex patterns among selected genes. Experimental results on real microarray datasets show that the proposed method outperforms the baseline methods with an accuracy of 99.2% and has been able to identify key genes effective in differentiating leukemia types by effectively reducing the data dimensions.

Economic and Productivity Impacts of Natural Gas in South Khorasan

Malihe Niksirat

Natural gas is a critical energy source that substantially contributes to meeting national energy demand. Iran possesses the world's second-largest natural gas reserves. In South Khorasan province, natural gas coverage is extensive, reaching 100% of urban households, 99.9% of rural households, and a large proportion of industrial facilities. This study examines the effects of natural gas supply to agricultural and production centers on production efficiency and economic profitability. The research is applied and employs a descriptive–survey design with a mixed-methods approach, combining structured questionnaires and expert interviews. The sample consisted of 165 respondents from the agricultural sector and 150 respondents from the production sector. Results indicate that supplying natural gas to agricultural and production centers enhances resource utilization, reduces energy costs, improves operational efficiency, and diminishes air pollution and greenhouse gas emissions.

Constrained Multi-Objective Deep Reinforcement Learning for Safe and Fair Urban Traffic Signal Control

Sara Motamed

This paper presents a constrained multi-objective deep reinforcement learning framework for urban traffic signal control. The problem is modeled as a constrained Markov decision process in which an agent simultaneously optimizes efficiency objectives while respecting explicit safety and fairness constraints. A dueling double deep Q-network (D3QN) is combined with a Lagrangian cost estimator to approximate both the reward value function and cumulative constraint costs. The state representation includes queue lengths, phase indicators and elapsed green times, and the action space consists of a small set of interpretable decisions such as extending the current green or switching to the next phase. The proposed controller is trained and evaluated in a SUMO-based microscopic simulation of a four-leg urban intersection under various traffic demand patterns. Its performance is compared with fixed-time, vehicle-actuated and unconstrained DQN controllers. Simulation results show that the proposed method can substantially reduce average delay and maximum queue length while keeping queue spillback and delay imbalance within predefined limits. These findings indicate that constrained multi-objective deep reinforcement learning offers a promising and practically deployable framework for safe and fair traffic signal control in congested urban networks, and can be extended to more complex corridors and network-wide settings in future work.

A spectral Conjugate Gradient method for solving unconstrained optimization problems

Mariya Toofan

The conjugate gradient method (CGM) stands out as one of the most rapidly growing and effective approaches for addressing unconstrained optimization problems. In recent years, significant efforts have been dedicated to adapting the CGM for tackling nonlinear optimization challenges. This research article introduces a new modification of the Fletcher–Reeves (FR) conjugate gradient projection method. The proposed method is characterized by its sufficient descent property, and its global convergence has been established under specific assumptions. Numerical experiments conducted on a range of functions from the CUTEr collection demonstrate the potential and effectiveness of the proposed methods.

An efficient solution algorithm for fuzzy linear fractional optimization problems with an application

Behrooz Alizadeh

In this paper, we investigate a solution procedure for a fuzzy linear fractional optimization problem in which the input parameters are considered as convex fuzzy numbers. By applying a specific fuzzy ranking method which is based on the α-cut concept, and according to Charnes and Cooper’s approach of variable transformation, the solution of the original fuzzy linear fractional optimization model is transformed to the solution of at most two semi-infinite linear programs that are dis similar among themselves via a sign in a constraint and in the objective function. An appropriate cutting plane algorithm(CPA) of Fang is uti lized to obtain the optimal solution of the semi-infinite linear programs. Further, the application of our provided algorithm in facility location theory is discussed properly. Finally, an illustrative example is given to clarify the developed solution procedure.

Intelligent detection of fraud in financial statements using deep learning and XGBoost

Mehdi Farrokhbakht

Fraud is a phenomenon that involves deviations and manipulations in financial statements. These actions can lead to tax non-compliance and erode the trust of investors and other stakeholders. Given the intricate nature and vast amount of financial data within organizations, leveraging artificial intelligence as a sophisticated tool can greatly enhance fraud detection in financial statements and bolster confidence in the face of evolving fraudulent tactics. Fraud or deception in the financial information of individuals or organizations reduces the level of trust and confidence that people have in the reliability and integrity of this information. This can lead to serious negative impacts, including loss of trust from customers, investors, and other entities, negative financial and legal consequences, and the exposure of illegal or improper operations that may involve financial crimes. This paper introduces an intelligent method for detecting fraud in financial statements. Initially, the Apriori algorithm is utilized to select pertinent features in the financial data. Subsequently, the performance of the proposed method is enhanced by augmenting the dataset using the GAN-CNN network. Finally, fraud detection is executed with the assistance of XGBoost. The results demonstrate that the proposed method has achieved a fraud detection accuracy of 95.3%.

Optimization inventory-related decisions under an outsourcing supply chain management using response surface methodology

Reza Ehtesham Rasi

The current research provides a mixed integer nonlinear mathematical programming model for a company that operates with several stores and multiple products, in which demand for each customer is characterized using fuzzy logic by triangular numbers, while the replenishment policy of each store for any product is the popular economic order quantity (EOQ) model under backorder. The throughput, dispatch, and budget constraints are considered in the proposed EOQ model. The objective is to integrate a vendor selection problem and EOQ policy, in which a multi-sourcing strategy is considered. In the proposed strategy, the ordered value of each store for any product can be split between one or more vendors. As such, a set of selected vendors can replenish each store for each product. This research aims to answer the following question as follows: (i) which vendors are chosen; (ii) which store is allocated to the selected vendors for each product; (iii) what is the optimal value for the inventory decisions.
The aim is to reduce the total cost of the company, including costs related to the vendor selection decisions along with the inventory decisions. To solve the mathematical model, a novel and practical genetic algorithm (GA) is developed then the response surface methodology (RSM) is utilized to tune its parameters. At the end, some numerical instances under different categories are evaluated to explain the applicability of the proposed approach.

Inverse data envelopment analysis for two-stage network systems with uncertain inputs and uncertain undesirable outputs

jafar pourmahmoud

Evaluation of healthcare systems, as a key organization providing different health services, is essential. This issue becomes more crucial when occurring crises such as a pandemic. They need to keep track of their success in the face of the crisis to assess the effects of policy changes and their capability to respond to new challenges. The inverse data envelopment analysis (InvDEA) technique is an applicable method in order to estimate the input/output levels of decision-making units (DMUs) to preserve predetermined technical efficiency scores. In classic studies of InvDEA, decision-Making Units (DMUs) as black boxes, ignoring their internal structure. This paper estimates input levels and new intermediate products to achieve a predetermined efficiency score set by the decision maker. In traditional inverse data envelopment analysis models, precise data are required to determine the input and/or output levels of each decision-making unit. However, in many scenarios, such as system flexibility, social and cultural contexts information may be indeterminate. In these cases, experts’ opinions are used to model uncertainty. Uncertainty theory, a branch of mathematics, logically deals with degrees of belief. This paper aims to develop an inverse Network DEA model incorporating uncertainty theory. We assume that inputs and outputs of decision-making units are based on experts’ belief degrees. To demonstrate the model is performance, we explore efficiency of healthcare systems during COVID-19 pandemic.

A ‎fuzzy multiobjective linear programming ‎problems for a supplier selection model under flexibility conditions

Roghayeh Yaser

Supplier selection is one of the main discussions in the Supply Chain. The issue of assigning purchase orders to suppliers that act differently in terms of quality, cast, services, etc. criteria is one of the significant concerns of purchase managers in the supply chain. To adopt an optimal decision in this regard is related to a multi-objective problem that the objectives are contradicting each other and have different importance and priority depending on the location. In practice, the existence of kind of ambiguity in explaining the information related to the problem constraints and complicated. In this regard, the emergence of Fuzzy set theory as a tool to describe such conditions besides presenting question model realistically can help to solve such problems well. Despite the importance of the model with the mentioned structure, unfortunately, few original works have been done in this field. As a result, in this paper, in addition to presenting a new multi-objective Fuzzy model being modelled based on assigning purchase order to suppliers in a supply chain a solution method is introduced based on using Fuzzy linear programming. To clarify solution process modelling and description, a case study is included related to selecting flour supplier for providing industrial bread of Khoshkar factory. The proposed model includes four objective functions:

Aggregate costs of minimizing type,
Services of maximizing type (such as packing, being faithful to promise, factory heath, discount, correct transportation, good relationships, honestly, etc.),
Flour useful survival of maximizing type (regarding monthly flour buying by the factory),
The purchased flour quality of maximizing type (concerning product type).

Especially in the solution process, a method is determined based on setting weight for each of the objectives concerning the major factory stockholders.

An Automated Stacking Framework for Insurance Customer Profitability Prediction using Hybrid Transformer-Gradient Boosting Architectures

Sohrab Kordrostami

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.

A sustainable supply chain network and optimizing on facility location and transportation network under uncertainty

Hadi Nasseri

One of the key challenges in supply chain management is the design of the supply chain network, which aims to determine the optimal locations of distribution centers across different regions in order to satisfy customer demand. In the proposed model, customer demand is fulfilled through distribution centers, which receive products from manufacturing plants. This study presents an integer linear programming model that simultaneously addresses supply chain network design and facility location decisions. The objective of the model is to minimize the total costs associated with establishing
distribution centers, transporting products from manufacturing plants to distribution centers, and distributing products from distribution centers to customers. To evaluate the effectiveness of the proposed model, several randomly generated test instances of different sizes were examined. Computational experiments were conducted using a linear programming solver and an iterative local search algorithm to compare their performance in obtaining optimal solutions. The results demonstrate that the iterative local search algorithm outperforms the linear programming solver by achieving optimal solutions with significantly shorter computational time across all tested instances.