MDP: Model-development Process

• Stage 1: Problem Formulation
  • Defining datasets that will be used for training and validation
  • Deciding which performance measures will be used for the evaluation of the performance of the final model
• Stage 2: Crisp Modelling
  • Focuses on the creation of first versions
  • Defines how complex the models needs to be
• Stage 3: Fine-tuning
  • Focuses on improving the initial version(s) of the model
  • Selecting the best one according to the pre-defined metrics
• Stage 4: Maintenance and Decommissioning
  • Monitoring the performance of the model after its implementation