The real-world applications increased demand for heterogeneous data classification for text, pictures, music, movies and medical data sets. The complexity of the learning class for an object that is associated with the set of values is a key issue for multi-value data sets. The present learning approaches are based on the features characteristics favouritism observed for similar sets of class values, but favouritism characteristics measured an object deviation value instead of association class. Such methods may be unfavourable for classification as each value is made up of specific features of characterization. However, few studies have tried to solved the problem through associating values learning over multi-value data sets. This paper presents a Multi-Value Association (MVA) Approach for efficient data classification based on associated pattern generation using binary multi-value association among the data. The objective of the proposal is to identify a Single-Class-Value (SCV) which can be most suitable along with the additional features patterns which can describe it most. To evaluate the efficiency we compare with some existing proposals using multi-value datasets which shows an improvisation.