LAMDA-HSCC: A semi-supervised learning algorithm based on the multivariate data analysis

• A semi-supervised learning algorithm based on a multivariate data analysis. • A new metric for semi-supervised contexts, called Semi-Supervised Criterion. • The first semi-supervised learning algorithm based on the LAMDA approach. • Migration, merging, and separation processes for assignment to classes/clusters. • Membership degree of an individual to a class/cluster based on features. In this work, we propose a semi-supervised learning algorithm, which can solve problems of classification, clustering, or a combination of them. This algorithm is based on the LAMDA family (Learning Algorithm for Multivariate Data Analysis), which computes the membership degree of an individual to a class or cluster considering the contribution of all features/descriptors. Thereby, it uses the LAMDA-RD approach for the clustering problem and the LAMDA-HAD approach for the classification problem. Also, it is composed of three sub-models for the migration, merging, and separation problems to improve the assignment of individuals to the classes/clusters. This proposal, called LAMDA- HSCC (Hybrid Scenarios of Classification and Clustering), is applied to several datasets of classification, clustering, and hybrid, in order to compare its performance with other algorithms, showing very encouraging results. Particularly, we define a new metric for evaluating performance in a semi-supervised context, called the Semi-Supervised Criterion (SSC), in which our approach achieves very good results.