Data Value Management – Mushroom Production Optimisation

Business Case

The development of a platform that uses machine learning models to identify key variables in mushroom production, providing real-time recommendations based on production status and predicting daily harvesting, optimising logistics and staffing.

Objectives

Identify the variables that most affected each stage of mushroom maturation, achieve real-time recommendations to maximise production, and predict daily harvesting to optimise logistics and staffing requirements.

Use case

In this project, historical data was structured, essential algorithms were trained, and an application was developed to visualise key variables for each phase. It also generates recommendations and displays predictions based on real-time production data. Finally, the system is retrained with the new data generated.

Infrastructure

On Premise

Technology

Machine learning and deep learning

Data

Variables associated with the incubated mushrooms’ characteristics (species, soil quality, fertilizers, pesticides, etc.), data on the atmosphere in which mushroom maturing takes place (humidity, CO2, temperature, etc.), irrigation, as well as the variables, target, quantity produced for each batch and daily quantity produced.

Resources

The resources used were the Data Value Management staff themselves who were in charge of preparing the data, implementing the predictive systems and training the users for the developed system. The entire development and deployment of the solution was carried out on the company’s servers.

Difficulties and learning

The primary challenge in this project was the fragmented data storage system, which spanned multiple systems with varying formats. Prior to project commencement, it was crucial to assist the company in defining databases and anticipating potential use cases and applications.

KPIs (business impact and metrics of the model)

The system resulted in a 6% increase in production and a 2% reduction in logistics costs. As the system is retrained and improved with new data, results are expected to improve in the future.

Funding

Collaborators, Partners

The project was carried out exclusively by DVM, with the help of the company’s staff who provided business knowledge.