Analysis Based On Linear Regression And SVM Leaf Quality And Level Of Predictive Models

Tobacco is the basic of cigarette industry, and has a pivotal role in the cigarettes quality. The flue-cured tobacco quality has four parts, including quality of tobacco leaf appearance, physical properties, chemical composition and sensory quality. The chemical compositions only evaluate some indicators of the tobacco, and the sensory indicators evaluate the pros and cons of tobacco intrinsic quality. Because the tobacco sensory quality evaluation mainly depends on the person’s physical and psychological conditions, coupled with the tobacco composition is more complex, making it difficult to establish determine the mathematical model between indicators. Introduction of computer intelligence technology in the tobacco industry can effectively solve these problems, can qualitative and quantitative analyze the tobacco quality, at the same time reduce the burden of manual evaluation.At present, the tobacco quality inspection and grading are based on the national grading standards, the GB is qualitatively express the tobacco characteristics, such as the national standard only on the quality of leaf appearance made a qualitative description, therefore, the tobacco grading process exist a certain degree of subjectivity and vagueness. In recent years, it is an important development trend of the application of artificial intelligence technology used in tobacco quality evaluation. This paper used the computer intelligence technology in the tobacco quality analysis and classification, and the papers specific work is as follows:(l)Firstly, this article have a brief analysis for the tobacco quality analysis content, as well as the main analysis methods, and the advantages and disadvantages of these various methods, and the problems in the field of tobacco quality analysis, the work of laying the groundwork for the thesis behind.(2) According to the large range of survey data, using tobacco samples data to make statistical analysis and description. Using PCA make data noise and dimension reduction by SPSS statistical analysis platform, and combined with linear regression analysis establish tobacco sensory quality prediction model, and then use the regression equation integrity inspection standards to verify the reliability of the model.(3) Using the tobacco chemical composition indicators, sensory quality indicators and tobacco leaf level as the sample data, combined with support vector machine algorithm that based on a small sample data processing establish tobacco leaf classification model in MATLAB software platform, and analysis the accuracy of the model and the predict results.(4) Using MATLAB GUI graphical user interface design tobacco quality analysis system, the tobacco quality analysis methods and process select through the menu, the analysis results display through the interface, it can be more intuitive seen tobacco quality and grasp tobacco quality and level objectively.This paper mainly studied the using of computer intelligence technology in the tobacco quality analysis and classification, the theory and method of principal component analysis and linear regression analysis, as well as the in-depth study of the key issues on support vector machine artificial intelligence technology, finally, attempt to guide the tobacco production and sell by the research methods and conclusions.