GateGroup-anjp

• 

  Interfaz

Inspiration

Food waste and ineffective stock rotation are actual issues in aircraft catering that we aimed to address. Flight attendants have a little idea of what is about to expire, and thousands of meals and beverages go bad every day before they are given. Our objective was to develop a system that guarantees improved demand planning, reduced waste, and better control.

What it does

We created an app that requires flight attendants to check and validate all products that are about to expire prior to every flight. First it needs a digital signature to confirm that they put those products in front of the cart; if not, they are unable to register sales.

Also, it helps company users to track each product's daily sales along with its amount, price, and optional photographs. Gives administrators access to a dashboard where they can monitor overall performance, review sales data, and view 30-day demand estimates for food and beverage categories based on artificial intelligence.

How we built it

Python and an app builder were used in the development of the application for the interactive dashboards and front end. Time-series forecasting using Statsmodels (SARIMAX), NumPy, and Pandas. Altair and Matplotlib are used to visualize data. GitHub + Visual Studio for experimentation and teamwork. Local CSV and Pickle files, kept in an artifacts folder, were used to manage all data.

Challenges we ran into

The primary obstacle was the scarcity of data, which initially caused the forecasts to be erratic. In order to prevent overestimation, we introduced a 67% adjustment factor that automatically lowers forecast results when variation or sample size is too low. In order to modify the algorithm to forecast both together and separately, we also had to properly combine and separate the food and beverage groups.

Accomplishments that we're proud of

Constructing a working prototype that incorporates the admin dashboard and real-time communication amongst flight attendants. Putting into practice an AI model that can adjust to low-data settings. creating a verification system that enforces signatures while maintaining traceability and operational discipline. Creating a simple, uncluttered dashboard that is understandable to non-technical people as well.

What we learned

We discovered that AI forecasting can be extremely effective when paired with human workflow design. Technology by itself is insufficient; the user experience must instill positive behaviors. Additionally, we learned how to clearly convey results to business customers who might not have a background in data and how to maximize predictions when dealing with tiny or noisy datasets.

What's next for GateGroup-anjp

Incorporate actual business data to develop more robust models. Allow for customized forecasts for each product (for example, view the 30-day forecast exclusively for Coca-Cola). To maximize stock per trip, incorporate automated inventory recommendations. Make the platform available from any airport worldwide by fully deploying it on the cloud.

Built With

• ml
• numpy
• pandas
• python