Maximum relevancy maximum complementary feature selection for multi-sensor activity recognition

• We propose a feature selection algorithm using MRMC.
• Show that MRMC provides a good result comparing to the 3 popular algorithms.
• The complementary measure improves the performance of the Clamping algorithm.
• Evaluate the proposed algorithm on 2 well-defined problems and 5 real life data sets.

In the multi-sensor activity recognition domain, the input space is often large and contains irrelevant and overlapped features. It is important to perform feature selection in order to select the smallest number of features which can describe the outputs. This paper proposes a new feature selection algorithms using the maximal relevance and maximal complementary (MRMC) based on neural networks. Unlike other feature selection algorithms that are based on relevance and redundancy measurements, the idea of how a feature complements to the already selected features is utilized. The proposed algorithm is evaluated on two well-defined problems and five real world data sets. The data sets cover different types of data i.e. real, integer and category and sizes i.e. small to large set of features. The experimental results show that the MRMC can select a smaller number of features while achieving good results. The proposed algorithm can be applied to any type of data, and demonstrate great potential for the data set with a large number of features.