Shortcuts to the project
A significant part of FMS's work involves entering customer orders into the transport management system (TMS) and creating shipments, with dispatchers using their expertise to fill in any missing information. The challenge was to use artificial intelligence to complete information from customer emails in such a way that at least the mandatory fields in the shipment entry form were filled in. Using statistical methods and machine learning, we were able to create a proof of concept that automatically completes missing information, thereby reducing the manual effort required and giving dispatchers more time for other tasks.
Our project approach
Data analysis of historical shipments
To achieve quick results, we limited the amount of data to a subset with the help of the customer's subject matter experts and performed a data analysis to better understand the data basis. In addition, we collected existing business rules that are applied when filling out the shipment forms and took special care to identify hidden rules. The aim was to reveal the intrinsic knowledge that experienced dispatchers apply unconsciously and that helps to complete individual fields.
Selection of features
The data had to be adapted and prepared for the application of machine learning models. To this end, a pipeline consisting of various data preparation steps was implemented. The pipeline ensures that the preparation of raw data into usable data is consistent. As the next step, we selected a suitable model from the field of machine learning for completing the shipments. This was implemented, evaluated, and improved in a test environment that was set up for this purpose.
Plausibility test of the results
The complete data set was used to complete the project. This phase focused intensively on evaluating and improving the results. To this end, we implemented another model and a benchmark for comparing performance. We adapted the models to the complete data set and revised the implementation of the data preparation pipeline. These optimizations further improved performance and were finally evaluated by our customer FMS.
Result
Incomplete shipment information can now be easily transferred to the system. Missing information is automatically added and returned as a complete data record. Intelligent models are used for this purpose, which are currently being tested in ongoing operations. In the future, clerks will receive technically appropriate suggestions and will be able to process shipments faster, more securely, and more efficiently.
Get in touch now
