Performance eva‎luation Mechanism for Organizations in Marketing: Integrating Data Envelopment Analysis, Neural Networks, an‎d the Balanced Scorecard (Case Study: E-Learning Platforms)

10/21/2026 12:00:00 AM

This study develops a hybrid framework for eva‎luating an‎d forecasting the performance of online-learning platforms. First, the four perspectives of the Balanced Scorecard—Learning & Growth, Internal Process, Customer, an‎d Financial—are modeled as a four-stage network Data Envelopment Analysis (DEA). Stage-level an‎d overall efficiencies are computed for 58 Iranian e-learning platforms over a five-year horizon (2019-2023), yielding 290 panel observations. Year-to-year Malmquist productivity indices are then derived, producing 232 progress/regress scores that identify 48 progressing an‎d 184 regressing platforms. To avoid the computational burden of solving four DEA models per decision-making unit (DMU) in each new period, a multilayer perceptron (MLP) is trained on 14 inter-period ratios (t + 1 / t) as inputs an‎d a binary progress/regress label (based on the Malmquist index) as output. The MLP, trained with the Levenberg–Marquardt algorithm on a 70 / 15 / 15 split of the dataset, achieves an overall classification accuracy of 97.4 %, with precision an‎d recall of roughly 97 % for both classes, while reducing execution time by more than 75 % compared with repeated DEA runs. These results show that the proposed BSC–DEA–ANN architecture provides a fast an‎d accurate tool for year-over-year productivity monitoring of e-learning platforms an‎d can inform strategic decision-making in this rapidly evolving industry.

كارت امتيازي متوازن , تحليل پوششي داده‌ها , شبكه عصبي مصنوعي , شاخص مالم‌كوئيست , آموزش مجازي

Balanced Scorecard , Data Envelopment Analysis , Artificial Neural Network , Malmquist Productivity Index , E-learning Platforms

Amirhossein khodaverdi

Mostafa Jafari