The reduction of polynomial degrees using moving average filter and derivative approach to decrease the computational load in polynomial classifiers The case study of feature extraction in carbon monoxide sensing application

Abstract

Carbon monoxide is a type of pollutant that is harmful to human health and the environment. On the other hand, carbon monoxide also has benefits for industrial matter. Since the benefits and disadvantages of carbon monoxide, the measurement of carbon monoxide concentration is required. The measurement of carbon monoxide level is not easy moreover with low-cost sensors. The usage of 4 sensors namely TGS2611, TGS2612, TGS2610 and TGS2602 has been used along with feature extractor. The polynomial classifier is required to interpret the feature vector into the amount of substance concentration. The common classifier methods suffer fatal limitations. The polynomial classifiers method offers lower complexity in solution and lower computational effort. Since the involvement of a huge number of data points in the modelling process leads to high degree in the polynomial model. The occurrence of Runge's phenomenon is highly possible in this condition. This phenomenon affects the accuracy level of the generated model. The degree reduction algorithm is required to prevent the occurrence of Runge’s phenomenon. The combination of MAF (Mean Average Filter) and derivative approach as degree reductor algorithm has succeeded in reducing the polynomial model degree. The greater the number degree in the model means the greater the computational load. The model degree reductor algorithm has been succeeded to reduce computational load by 96.6%.