eva‎luation an‎d Prediction of Efficiency in a Food Industry Holding in a Long-Term Horizon Using an Integrated Approach of Data Envelopment Analysis an‎d Machine Learning: A Case Study

1/31/2027 12:00:00 AM

The food industry is one of the fundamental pillars of national an‎d global economies, playing a vital role in ensuring food security, supporting sustainable development, an‎d enhancing quality of life. This sector faces multiple challenges, including economic fluctuations, intensified competition, shifting market deman‎ds, an‎d environmental requirements—factors that underscore the necessity of eva‎luating an‎d optimizing the performance of holding companies operating in this domain. This study aims to assess an‎d predict the relative efficiency of the subsidiaries of a food industry holding in Iran over a seven-year period from 2019 to 2025. The research adopts an integrated approach combining Data Envelopment Analysis an‎d machine learning algorithms. In the first stage, the SBM-UO model is applied within the DEA framework to measure relative efficiency while incorporating both desirable an‎d undesirable outputs (such as waste an‎d pollutants). This enables the identification of inefficiencies an‎d improvement opportunities in resource utilization an‎d in generating desirable an‎d undesirable outputs. In the second stage, the calculated efficiency scores are used as the basis for labeling the units (efficient/inefficient), an‎d the resulting dataset serves as the input to machine learning algorithms. To predict future efficiency an‎d classify units over the long term, Artificial Neural Networks (ANN) an‎d Classification an‎d Regression Trees (CART) are employed. ANN enhances predictive accuracy by modeling complex nonlinear relationships, while CART generates hierarchical decision paths, offering interpretability an‎d managerial insight into the behavioral patterns of units.Input an‎d output data were collected from credible sources including annual reports, financial statements, an‎d market information, an‎d were preprocessed for efficiency analysis an‎d machine learning training. The findings of the study not only clarify the relative efficiency status of the subsidiaries an‎d distinguish efficient from inefficient units but also identify key factors influencing efficiency. Furthermore, the research presents an integrated framework combining SBM-UO with ANN an‎d CART, which can be used to optimize resource allocation, reduce costs, enhance productivity, an‎d support strategic decision-making within food industry holdings. This framework may also be generalized for eva‎luating an‎d predicting performance across other companies in the sector.

تحليل پوششي داده‌ها , صنايع غذايي , كارايي , شبكه عصبي مصنوعي , درخت تصميم‌گيري

Data Envelopment Analysis , food industry , efficiency , artificial neural networks , decision trees

Reyhane Vahabpour

Hadi Sahebi