| Resumen |
In this paper the Naïve Associati ve Classifier (NAC), a novel supervised learning model, is presented. Its strengths lie in its simplicity, transparency, transportability and accuracy. The creation, design, implemen- tation and application of the NAC are sustained by an original similarity operator of our own design, the Mixed and Incomplete Data Similarity Operator (MIDSO). One of the key features of MIDSO is its ability to handle missing values as well as mixed numerical and categorical data types. The proposed model was tested by performing numerical experiments using finance-related datasets including credit assignment, bank telemarketing, bankruptcy, and banknote authentication. The experimental results show the ade- quacy of the model for decision support in those environments, outperforming several state-of-the-art pattern classifiers. Additionally, the advantages and limitations of the NAC, as well as possible improve- ments, are discussed. |
