Due to NDA, the project content shown at the expo is limited. We can't provide any documentation for the project and any information regarding the given system and project partner. As the given network system contains the priorities and different parameters for each package, the spikes in the traffic at peak time have slowed down the network system. We have developed a machine learning model to lower the operating cost and transport time of the package by analyzing and adapting the changing traffic of the packages. The model is developed with Pytorch, and API (Application programming interface) is developed with C++ 11 and C++ libraries. The model first fetches the training dataset from the database provided by the system. Then the model is trained with the gradient descent method and provided training dataset. After the model has been trained, the model receives the actual package data from the system through the socket. The model looks at the different parameters of the package and reorganizes the priority of each package. This modified data set is sent back to the system through the socket. Finally, the system uses modified data set to provide faster and reliable network traffic at any time. Based on the gradient descent method, the model should output a convex function behavior regarding priority and wait time. The model should reduce the wait time of each packet with the increase of the transaction. However, the model sometimes increases the wait time with the increase in transaction due to the error. Although the current model is not implementable yet, it provided a good framework for another group to have a more stable footing to make a functional and implementable model. With the improvement of the model, society can experience a faster and more reliable network system as the system changes the package's transportation to fit the traffic best.
