Building a self-learning forecasting logistics system that allows you to build a variety of schemes for the workload of the hub, build a scheme of calls of vessels at the port and their sequence. The task within the project is divided into four stages.
To solve the problems of classification and regression, various methods were used, including the logistic regression method, the selected regressors were loaded into the recursive neural network LSTM, and a system was constructed from several recurrent networks.
C ++, XGBoost, Python (Pandas, Scikit-learn, Keras libraries)
71% of the project was completed (three of four stages).