Study On Application Of Discriminatory Analysis And Logistic Regression Model In Crude Oil And Fuel Oil Classification

In order to meet the needs of the national economic development, our countryhas to import a mass of crude oil and fuel oil from the international market every year.At present, imports of crude oil and fuel oil were implemented in different policiesthat a strict quota system which only a few of large enterprises have importqualifications is applied in the imports of crude oil and an automatic licensing systemis applied in the imports of fuel oil imports in China. Both the crude oil and fuel oilare superficial similarity with the property of black or dark brown viscous liquid.Therefore, some traders import crude oil in the name of the fuel oil. This behavior is akind of smuggling. However, our country investigates merely cases and lack offull-scale, systemic, dynamic supervision for this kind of behavior. In order toeffectively combat the behavior of smuggling, it seems to urgency establishing aprediction model which is classification of crude oil and fuel oil day by day.In this study, a group of fifty-four samples (thirty crude oil samples andtwenty-four fuel oil samples)coming from different fields around the world has beencollected to detect n-alkanes(n-C7~n-C30), pristine (Pr) and phytane (Ph) by gaschromatography coupled to mass spectrometry(GC-MS), uses discriminatory analysis,logistic regression analysis methods and principal component analysis(PCA) andother methods, uses SPSS statistical analysis software, creates prediction models forthe identification of crude oil and fuel oil. This paper also does comprehensiveanalysis and evaluation of the results of applied analysis, gets a review of someproblems existed in this study. The classification prediction model of crude oil andfuel oil established in this paper not only can realize the full-scale, systemic, dynamicsupervision, but also can provide a guide for the oil spill work.The conclusions in this thesis are included as follows: 1. The discriminatory analysis model and logistic regression analysis model ofcrude oil and fuel oil species prediction are established. Research found that both thediscriminatory analysis, logistic regression analysis can be accurately predicted theclassification of crude oil and fuel oil, and the logistic regression analysis predictionresult is better than the discriminatory analysis. The discriminatory analysis modelincludes bayes discriminatory analysis and fisher discriminatory analysis, theaccuracies of this two discriminatory analysis models are both88.9%; and theaccuracy of logistic regression analysis model is100.0%, the goodness of fitness testresults show that the logistic regression analysis model can well explain the fact data.2. This study combines the principal component analysis (PCA) with thediscriminatory analysis and logistic regression analysis, establishes the PCAdiscriminatory analysis model and PCA logistic regression analysis model. Theaccuracies of the PCA bayes discriminatory analysis model and the PCA fisherdiscriminatory analysis model are both88.9%; and the accuracy of PCA logisticregression analysis model is83.3%. The results show that prediction results of thePCA discriminatory analysis model and PCA logistic regression analysis model areless better that the general discriminatory analysis model and logistic regressionanalysis model.The innovation in this thesis is included as follows:In this thesis, chemical pattern identification of classification prediction of crudeoil and fuel oil is first used, and first realizes the classification identification of crudeoil and fuel oil completely bases on chemical index. Taking n-alkanes (n-C7~n-C30),pristine(Pr), phytane(Ph) as observation variables, uses multiple statistical analysis astheoretical guidance, uses SPSS statistical analysis software, establishesdiscrimination functions for crude oil and fuel oil classification prediction, and putsforward the discrimination criterions. Find a way out of the difficulty that our countryinvestigates to the smuggling behaviors imports crude oil in the name of fuel oilmerely to cases and lacks of full-scale, systemic, dynamic supervision. Overcome theinsufficiencies of traditional crude oil and fuel oil classification identifications whichmainly base on physical parameters with empirical values.