6.3 The Steps of Building a Model

flowchart tab1 Data Collection Ensure data completeness tab2 Data Cleaning Handle missing values, remove outliers tab1->tab2 tab3 Feature Selection Select relevant variables tab2->tab3 tab4 Model Training Apply chosen algorithms tab3->tab4 tab5 Model Evaluation Assess model performance tab4->tab5 tab6 Model Tuning Optimise model parameters tab5->tab6 tab7 Model Deployment Integrate into real world applications tab6->tab7

The Model Function

y=f(x1,x2,...)+ϵ

where

y is the response, e.g., number of deaths or infections.

x1,x2,... are predictors that might affect y, e.g., vaccinations, demographics.

ϵ is the error term (difference between observed values of y and predicted values from the model.

  • Machine learning algorithms optimise the coefficients to minimise the difference between the observed and predicted values, and can handle complex structures.

  • Model function can be used on new x values to predict future values of y.