The idea is to have a Machine Learning model capable of predicting when each machine will need maintenance and thus avoid unscheduled downtime.

But before using a predictive model, top management needs to understand how the model makes predictions and which metrics have the greatest impact on the model's prediction.

• Productivity (predictor variable)
  • This is a measure of productivity, which describes the amount of time a machine is working at peak performance. The metric is a product of machine availability, performance and quality.
• Performance (predictor variable)
  • This is the portion of products that come off the production line without defects and meet specifications without the need for any grinding work.
• Cost (predictor variable)
  • This is the cost of electricity, steam, oil or gas needed to produce a given unit of product in the factory.
• Priority (predictor variable)
  • Equipment priority when entering the maintenance period (Low, Medium, High).
• Efficiency (predictor variable)
  • The amount of product a machine produces during a specific period. This metric can also be applied to the entire production line to verify its efficiency.
• Maintenance (target variable):
  • 0 means the equipment does not require maintenance (No)
  • 1 means the equipment requires maintenance (Yes)