4.7 Weights

This option allows you to assign individual weights to each instance by choosing a special weight field. This is useful when you have an unbalanced dataset, where data instances ofa given kind, e.g., those indicating a fraudulent transaction, are scarce in comparison to other ones. In such case, you may want to assign more weight to the scarce instances so they are equivalent to the abundant ones.

The selected field must be numeric, and it must not contain any missing values. The field selected for weighting purposes will not be taken into account as an input to calculate cluster distances. You can select an existing field in your dataset, or you can create a new one to assign customized weights.

For example, below you can find a transactional dataset example for which we included a field called “Weight”. This field indicates that fraudulent instances weigh 10 more times than non fraudulent ones. BigML Flatline editor is a powerful tool for adding new fields to your dataset, such as a weight field. Another field that could be used in this example may be the transaction “Amount” so that transactions with higher amounts will have higher weights in the cluster.