چکیده:
One of the models that is used to measure the efficiency and decision support is DEA, but considering all the merits, it also has its own limitations, which has led to the idea of combining it with artificial neural networks. ANN is increasingly used in model and data-based approaches to enrich analytical and predictive capabilities and thus improve decision-making. The present research presents a model for evaluating the efficiency of units by integrating neural networks. The studied sector is the active pharmaceutical industry in the Tehran stock market. To create the model, the efficiency of 4 DEA models on a variable scale, including input-oriented and output-oriented BCC model, SBM model and RAM model during the years 2018 to 2022 was calculated in GAMS. The efficiency values of these four models were ANN'S education vector. Also cost, income and profit data from 2018 to 2022 was entered into MATLAB as ANN'S input. To generalize education, the data of 2023 was used. The results showed that the trained efficiency boundary shows a more comprehensive and accurate approximation of efficiency for the ranking of pharmaceutical companies. The results of this research will help pharmaceutical companies in the fields of investment, resource allocation, predicting the results of policies and planning.
خلاصه ماشینی:
Presentation of a hybrid model of Data Envelopment Analysis and Artificial Neural Networks for ranking the efficiency of pharmaceutical companies 2 Mostafa Ebrahim Pour Azbari *1, Aida Fallahpour Mobaraki 1.
To answer the above questions and fill this research gap, this study attempts to simulate the efficiency of decision-making units using Variable Returns to Scale Data Envelopment Analysis models such as BCC input-oriented, output-oriented, SBM model, and RAM model.
In this manner, to obtain a continuous efficiency frontier in variable returns to scale, the efficiency data of four Data Envelopment Analysis models, including input-oriented BCC, output-oriented BCC, SBM model, and RAM model, were provided as training vectors to the neural network.
In this research, using the output-oriented CCR model, the efficiency of units in 2011-2012 was obtained, and after training the network, the industrial parks were ranked based on 2012-2013 data and the network's prediction capability [22].
Ajli and Safari (1390), referring to the weakness in the discriminatory power of units in Data Envelopment Analysis, conducted a study to rank gas companies in 23 provinces using the multiplicative CCR model and the Anderson-Peterson (AP) method 1, and finally combined it with artificial neural networks to overcome the weakness of Data Envelopment Analysis, and then used the hybrid NN-DEA model to predict the performance of gas companies.
3- Research Methodology The objective of this research is to rank pharmaceutical companies using a hybrid model of Data Envelopment Analysis (DEA) and Artificial Neural Networks in terms of variable returns to scale efficiency.
In this research, a hybrid model of Data Envelopment Analysis and Artificial Neural Networks was presented for ranking the efficiency of pharmaceutical companies.
(
, 
, 
)
