Logistic regression, also known as nominal regression, is a statistical technique for classifying records based on values of input fields. It is analogous to linear regression but takes a categorical target field instead of a numeric one. Both binomial models (for targets with two discrete categories) and multinomial models (for targets with more than two categories) are supported.
Logistic regression works by building a set of equations that relate the input field values to the probabilities associated with each of the output field categories. Once the model is generated, it can be used to estimate probabilities for new data. For each record, a probability of membership is computed for each possible output category. The target category with the highest probability is assigned as the predicted output value for that record.
Binomial example. A telecommunications provider is concerned about the number of customers it is losing to competitors. Using service usage data, you can create a binomial model to predict which customers are liable to transfer to another provider and customize offers so as to retain as many customers as possible. A binomial model is used because the target has two distinct categories (likely to transfer or not).
Multinomial example. A telecommunications provider has segmented its customer base by service usage patterns, categorizing the customers into four groups. Using demographic data to predict group membership, you can create a multinomial model to classify prospective customers into groups and then customize offers for individual customers.
Requirements. One or more input fields and exactly one
categorical target field with two or more categories. For a binomial model the target must have a
measurement level of
Flag. For a multinomial model the target can have a
measurement level of
Flag, or of
Nominal with two or more
categories. Fields set to
None are ignored. Fields used in
the model must have their types fully instantiated.
Strengths. Logistic regression models are often quite accurate. They can handle symbolic and numeric input fields. They can give predicted probabilities for all target categories so that a second-best guess can easily be identified. Logistic models are most effective when group membership is a truly categorical field; if group membership is based on values of a continuous range field (for example, high IQ versus low IQ), you should consider using linear regression to take advantage of the richer information offered by the full range of values. Logistic models can also perform automatic field selection, although other approaches such as tree models or Feature Selection might do this more quickly on large datasets. Finally, since logistic models are well understood by many analysts and data miners, they may be used by some as a baseline against which other modeling techniques can be compared.
When processing large datasets, you can improve performance noticeably by disabling the likelihood-ratio test, an advanced output option.