The nature of statistical learning theory
The nature of statistical learning theory
An overview of statistical learning theory
IEEE Transactions on Neural Networks
| Hi-index | 0.00 |
Since non-linearity is obviously characteristic of natural gas analysis, and few ideal spectrum data samples can be actually obtained from the mass natural gas, the accuracy level of concentrating each component in the natural gas turns out to be far from high. In response to the dilemma above, a multi-level- and SVM-subset- based infrared spectrum analyzing method is proposed for the analysis of natural gas. According to the idea of natural gas distribution pattern recognition→natural gas analysis ←result output, the new analyzing method, as based on multi-level and SVM subset, consists of two levels: the pattern recognition level as well as the analysis and the result output level. The pattern recognition level serves to identify the natural gas distribution pattern, whereas the analysis and the result output level is the concrete natural gas component concentration analysis and the result output level, with the established SVM calibration model designed to analyze and calculate the natural gas component concentration. The experimental results show that the component concentration maximal deviation is 0.49% and maximal average deviation is 0.059%. The method can work for other natural gas infrared spectrum analyses, and therefore has the theoretic and application value.