2.2 Input Data for Anomalies

BigML anomalies support categorical and numeric fields as input fields. If the dataset contains text and items fields, they will be automatically included in the model as ID fields, but they will not be used to compute the anomalies. (See section 4.4 .)

Moreover, anomalies include missing values as valid values in the computation by default. Consequently, instances with missing values may appear among the top anomalies if anomalous patterns can be learned from missing data. By using the BigML API you can replace missing values so they are not considered by the model.