In semi-supervised classification we work with both unlabeled and labeled examples. Two main settings can be found: transductive and inductive classification. The former concerns the problem of predicting the labels of the unlabeled examples, given in advance (in the training set), by taking both labeled and unlabeled data together into account to train a classifier. The latter considers the given labeled and unlabeled data as the training examples, and its objective is to predict unseen data.

All data sets included here are in the section containing Standard data sets. All of them, which were partitioned using a 10-fold cross-validation procedure, were employed in the paper:

I. Triguero, S. García, and F.Herrera, Self-Labeled Techniques for Semi-Supervised Learning: Taxonomy, Software and Empirical Study. Knowledge and Information Systems, 42 (2015) 245-284.

It is noteworthy that test partitions are kept aside to evaluate for inductive purposes.

The training partitions have been divided into labeled and unlabeled examples. In the division process we do not maintain the class proportion in the labeled and unlabeled sets since the main aim of semi-supervised classification is to exploit unlabeled data for better classification results. Hence, we use a random selection of examples that will be marked as labeled instances, and the class label of the rest of the instances will be removed.

Four labeled ratios are considered: 10, 20, 30 and 40 %. Thus, for instance, assuming a data set that contains 1,000 examples, when the labeled rate is 10 %, 100 examples are put into the labeled set with their labels, while the remaining 900 examples are put into the unlabeled set without their labels.

This section shows the semi-supervised classification data sets avalaible in this repository. Every one defines a semi-supervised classification problem, where each of its examples is composed by some nominal or numerical attributes and a nominal output attribute (its class).

These data sets are composed of three files for each partition: training, transductive and test partitions. The former is composed of labeled and unlabeled instances (labeled as ''unlabeled''). Transductive partition contains the real class of unlabeled instances and the latter collect the test instances.

Each data file has the following structure:

• @relation: Name of the data set
• @attribute: Description of an attribute (one for each attribute)
• @inputs: List with the names of the input attributes
• @output: Name of the output attribute
• @data: Starting tag of the data

The rest of the file contains all the examples belonging to the data set, expressed in comma sepparated values format. None of the data sets contains missing values.

We offer information about experimental studies using these data sets (result files, papers and more) in the Experimental studies with Semi-supervised classification data section of the repository.

Four labeled ratios are considered:

 -    10% of labeled data
 -    20% of labeled data
 -    30% of labeled data
 -    40% of labeled data