Abstract:
"
Courier shipment industry is widely popular in the international trade due to the fact that anyone
could order something over internet and receive the shipment at their doorstep without breaking a
sweat. Clients expect their packages to be received at the same day they shipped to the country.
Successful courier services who have more than enough delivery resources can fulfill the client’s
expectations without a problem. Yet, the small scale courier services and startups are more than
happy to manage their limited delivery resources to provide for the same customer satisfaction to
grow their business. The purpose and the importance of this research is to find an efficient
method for the courier services with limited resources on how to manage their delivery resources
efficiently for better business.
Machine learning, one of the most significant components in artificial intelligence, enables
machines to self-learn from data and perform tasks without human intervention or explicitly
programming. Vary in type, machine learning could be employed in many strategies. Time series
forecasting is the process of observing the particular time intervals in historical values to extract
meaningful statistics. A time series based machine learning model can provide the future
predictions by analyzing a dataset. This dissertation is a result of the project to implement a time
series based machine learning pipeline to predict the shipment orders received to a courier service
and allocate resources for the distribution of the shipments, based on the prediction.
The machine learning pipeline provides time series based machine learning model implemented
on LSTM (Long Short-Term Memory networks) algorithm to forecast future predictions.
Predicting the shipment trend of shipments a courier service receives in order to allocate the
delivery resources and driver allocations are results produced by this dissertation. Compared to
the similar systems existing but belongs to different domains, it performs equally or better. The
system is available as an API which could be integrated with exiting operative courier systems,
and as per the testing purposes, a web graphical interface is facilitated for the user before taking
decisions on integration."
