A Multicriteria Approach based on Rough Set Theory for the Incremental Periodic Prediction

Abstract This paper proposes a Multicriteria Approach for the Incremental Periodic Prediction (MAI2P). This approach is periodically applied while considering the sequential evolution of the dynamic information system under the variation of the set of actions in an ever-evolving learning sample. It is based on the Dominance-based Rough Set Approach (DRSA) and consists of three phases. The first aims at constructing a decision table and is based on three steps: (1) constructing a representative learning sample of “Actions of reference”, (2) constructing a coherent criteria family for the actions’ characterization and (3) building a decision table. The second consists in an incremental updating of the DRSA approximations in order to infer a preference model resulting in a set of decision rules. The third consists of classifying the potential actions in one of the predefined decision classes. The first two phases run at the end of the current period and the third phase runs at the beginning of the next period. The approach MAI2P has been applied in the context of Massive Open Online Courses (MOOCs). It has been validated on a French MOOC proposed by a Business School in France. Experiments showed that the pessimistic cumulative approach gives the most efficient preference model with an F-measure and an accuracy values reaching 0.66 and 0.89 respectively.