A novel error-correcting output codes based on genetic programming and ternary digit operators

•A novel genetic programming (GP) based ECOC algorithm is proposed with the aid of ternary digit operators.•In the GP, every terminal node represents an ECOC column along with a feature subset, and every nonterminal node is a ternary digit operator.•Each individual is transformed into a codematrix associated with a set of feature subsets.•A local optimization algorithm is proposed to remove redundant or irrelevant columns and append necessary columns to tackle hard classes.•TOGP-ECOC outperforms other well-known ECOC algorithms on various data sets. The key to the success of an Error-Correcting Output Code (ECOC) algorithm is the effective codematrix, which represents a set of class reassignment schemes for decomposing a multiclass problem into a set of binary class problems. This paper proposes a new method, which uses Ternary digit Operators based Genetic Programming (GP) to generate effective ECOC codematrix (TOGP-ECOC for short). In our GP, each terminal node stores a ternary digit string, representing a column and a related feature subset; each non-terminal node represents a ternary digit operator, which produces a new column based on its child nodes. In this way, each individual is interpreted as an ECOC codematrix along with a set of corresponding feature subsets, serving the solution for the multiclass classification task. When a new individual is produced, a legality checking process is carried out to verify whether the transformed codematrix follows the ECOC constraints. The illegal one is corrected according to different strategies. Besides, a local optimization algorithm is designed to prune redundant columns and improve the performance of each individual. Our experiments compared TOGP-ECOC with some well known ECOC algorithms on various data sets, and the results confirm the superiority of our algorithm. Our source code is available at: https://github.com/MLDMXM2017/TOGP-ECOC.