What type of classification attempts to group cases based on similarities in input variables without prior knowledge of outcomes?

Unsupervised classification is the appropriate choice because this method focuses on grouping data based solely on the features or attributes present in the dataset, without having any prior knowledge of the outcomes or labels associated with those groups. This technique is essential in exploratory data analysis where the goal is to identify patterns, similarities, or underlying structures within the data itself.

In contrast, supervised classification relies on pre-existing labels linked to the data, guiding the model to learn from examples before predicting outcomes for new cases. Regression classification, while it may seem relevant, typically involves predicting a continuous outcome rather than classifying into groups based on input variables. Hierarchical classification organizes data into a structured hierarchy but also requires knowledge of the outcomes to establish those connections.

Thus, unsupervised classification stands out as the method that fits the criteria of grouping cases by their similarities without any predefined outcomes, making it the correct answer.