Context
In the logistics domain, one of the critical Key Performance Indicators (KPIs) is tonnage the total weight of cargo handled over a specific period. The challenge the business team faced was that their tonnage forecasts were largely based on intuition or domain expertise rather than data-driven methods. This approach often led to inaccurate predictions, affecting decision-making and resource allocation. To address this, there was a need to develop data-driven forecasting models that could accurately predict tonnage for a given horizon. The client required forecasts at multiple levels of granularity: airport level, cost center/building level, customer/airline level, and commodity level. As the granularity increases, so does the complexity of the forecasting models. Therefore, it was essential to identify the various factors or features that could influence tonnage and determine whether these features were available in the database for modeling. This project is ongoing, and as it progresses, the technologies to be used will be determined based on the needs identified during development.
Requirements
Data-Driven Tonnage Forecasting:
Develop accurate forecasting models for tonnage based on historical data, moving away from guesswork and intuition.
Provide forecasts at multiple levels of granularity, including airport, cost center/building, customer/airline, and commodity levels.
Feature Identification and Selection:
Collaborate with domain experts to identify potential features that could influence tonnage forecasts.
Ensure that identified features are available in the existing database or can be derived from available data.
Model Complexity Management:
Manage the increasing complexity of models as the forecasting moves from broader levels (airport) to more granular levels (customer/airline, commodity).
Ensure that the models are scalable and can handle the complexity of multi-level forecasting.
Ongoing Development and Adaptation:
Adapt the models as new data becomes available or as additional features are identified.
Maintain flexibility in the approach to accommodate the evolving needs of the project and the discovery of new insights during development.
Approach
Ongoing Development:
The project is ongoing, with continued collaboration between data scientists and domain experts to refine models and incorporate new insights.
Technologies and tools will be selected as the project progresses, ensuring that the chosen solutions align with the project’s evolving needs.
Technologies Used
The specific technologies will be determined as the project progresses and based on the needs identified during development.