Saydulloev Inomjon Inoyatovich

Annotation: 

The article discusses the application of predictive modeling methods in the Green Logistics module, a software platform developed for Mavsim and Co LLC in order to increase the environmental sustainability of logistics processes. The author explores the effectiveness of using logistic regression for binary classification tasks related to route optimization, emissions forecasting, and environmental assessment of transport operations. A comparison of logistic regression with linear, ridge, and LASSO regression has been carried out, which allows us to identify differences in cost functions, model assumptions, and application features. Based on the analysis, a hybrid approach combining logistic regression and artificial neural networks is proposed. This approach uses logistic regression coefficients as initial weights to improve the recognition of nonlinear dependencies and improve the accuracy of forecasts. The results show that the hybrid model provides an 8-10% increase in accuracy compared to traditional methods, especially in low-emission route identification tasks. The article also discusses evaluation metrics (accuracy, precision, recall, F1-score) that confirm the effectiveness of the proposed solution. The presented approach contributes to the development of sustainable logistics, improved planning of transport operations and reduction of negative environmental impacts.

Keywords: green logistics, predictive modeling, logistic regression, LASSO, ridge regression, neural networks, hybrid models, route optimization, emission reduction, sustainable transport operations.