Gas Identi cation by Using a Cluster-k-Nearest-Neighbor

Abstract. Among the most serious limitations facing the success of future consumer gas identification
systems are the drift problem and the real-time detection due to the slow response of most of todays gas
sensors. In this paper, a novel gas identification approach based on Cluster-k-Nearest Neighbor. The
effectiveness of this approach has been suc-cessfully demonstrated on an experimentally obtained data set.
Our classify takes advantage of both k-NN which is highly accurate and K-means cluster which is able to
reduce the time of classification, we introduce Cluster-k-Nearest Neighbor as “variable k”-NN dealing with
the centroid or mean point of all subclasses generated by clustering algo-rithm. In general the algorithm of Kmeans
cluster is not stable in term of accuracy. Therefore for that reason we develop another algorithm for
clustering space which contributes a higher accuracy compares to K-means cluster with less subclass number,
higher stability and bounded time of classification with respect to the variable data size. We find 98.7% of
accuracy in the classification of 6 different types of Gas by using K-means cluster algorithm and we find
almost the same by using the new clustering algorithm.