Predictive Modeling of Eco-Friendly Logistics using Random Forest

The Conventional supply chain frameworks typically focus on cost and speed while relegating environmental metrics such as carbon emissions and energy consumption to the background. By doing so, they hinder truly sustainable operational choices. The current research confronts this limitation head-on by introducing a Random Forest-Based methodology that merges ecological and logistical variables for greener freight optimization. To train the model, the authors created a synthetic dataset containing 5,000 transaction records. The dataset was enriched with features including carbon output per mile, product lifecycle footprint, and expected delay risk, all of which help classify more sustainable routing and handling options. Pre-Processing steps employed StandardAero to normalize continuous measures and OneHotEncoder to transform categorical entries, ensuring statistical balance and predictive accuracy. The final Random Forest architecture consisted of 200 individual trees, each limited to a depth of 10, and delivered impressive performance metrics: 91.1% overall accuracy, precision and recall values both exceeding 0.89, an F1-score of 0.905, root-mean-square error (RMSE) of 0.255, and an R-squared of 0.81. These figures collectively demonstrate the model’s durability and trustworthiness in spotting environmentally preferable paths. Because it is both scalable and interpretable, the framework acts as a practical decision-support vehicle for embedding sustainability into day-to-day logistics planning.