Acheter le livre

10h11 won the HACKATHON Geodis Smart Warehouse

Discover the reasons that led our team to come to this challenge and how this operation went.

On November 8, 2018, 10:11 a.m. participated in the Geodis Smart Warehouse Hackathon and won the prize dedicated to data-notification. On this occasion, we wanted to share this experience with you, explaining in detail the work that has been done in data science and datavisualization.

Why did we decide to participate ?

First of all, it is one of the few HACKATHON that offered a datascience and datavisualization mix, i.e. the ability to predict and interface data in the same service. This dual objective being similar to the demands of our customers as well as our know-how, we have decided to provide an answer to this hackathon.

How did the HACKATHON go ?

We were able to obtain a data set upstream of the HACKATHON to start building our scientific approach. The subject: predicting the entry and exit of Geodis warehouses. In parallel, we imagined a service that would allow us to visualize the results, but also to ensure that the use of the service natively builds a database in real time. Thus, the tool becomes a technological asset that increases in value through its use.

Then, we had 24 hours with Geodis to present our concept, our first results and adjust the whole to win the HACKATHON.

How did the HACKATHON go?

  • A predictive engine

To build the Geodis warehouse entry and exit prediction engine, we deployed a scientific methodology based on a daily preparation history that we consolidated. We considered 4 main variables on which we applied feature engineering, i.e. the design of a new variable from the variables present in the original database. The initial parameter was the prediction period according to the customer’s request and our performance criterion RMSE (Root Mean Squared Error).

  • A visualization service

To help warehouse staff, we have designed a mobile app to visualize the inputs and outputs predicted by the predictive engine. In parallel, the app allows you to adjust the storage areas of each programmed entry or exit according to its rotation index (the rotation index is the index used to define the exit frequency of a package on a pallet. The higher the index, the closer the pallet concerned must be to the exit areas of the warehouse). The objective here is to limit the movements of each person in the warehouse and to optimize the management of the location of stocks by the actual use made of them. Finally, the app becomes “intelligent” because it allows you to quickly store each entry and exit operation from the warehouse. Through this system, the rotation index is updated in real time, the app learns from its use via the predictive engine based on “machine learning”. The tool then makes it possible to adjust warehouse usage by understanding parcel rotations in near real time. The basic history is plugged to an algorithm that proposes uses. The service, through its use, then creates again a history that allows the algorithm to learn and improve the future use: a virtuous circle.

Vimeo link :

In concrete terms, what form does our app and engine take ?

The engine performance report provides package (pallet) predictions with an average deviation of +/- 15% between the actual and predicted number of packages (pallets). These results are based on a first version of the engine. Optimization is highly conceivable, particularly through parameter optimization.

10h11 is convinced that the service, through use, creates as much value as statistical modeling (AI). It is the complementarity of the two that makes the added value of tomorrow’s technological projects. It is with this objective in mind that we deal with all the issues that our clients entrust to us. Whether for Coca-Cola (see case study) or Suez (see case study), our aim is to serve to facilitate the use of technologies through data in a search for performance optimization.

Victory at this hackathon is one thing and it makes us proud essentially by the fact that we have met Geodis’ needs. Supporting the industrial world in its relevance is a powerful leitmotiv for our entire team.

0 / 5 0
Cart Overview