基于商空间的黄金价格SVM模型预测

Based on the detailed analysis of the influencing factors of the gold price，the gold price of China was predicted according to the price factor of gold price combined with the quotient space theory and the support vector machine method.Firstly，we used the person correlation coefficient method to compare the correlation between 9 price factors with gold price in the current.Five price factors with larger correlation coefficient were selected，they are US dollar index，WIT crude oil futures，G5 currency index，producer index，consumer index and correlation coefficient with correlation coefficient of -0.46，0.52，0.5，0.59 and 0.55 respectively.Secondly，through the granger causality test，we got the reason that the commodity index and the consumer price of the index lead to the change of the gold price hypotheses are more likely to be established，the odds are 0.83338 and 0.95609 respectively.Then，using the dollar index，the WIT crude oil futures，the G5 currency index，the producer index，the consumer index，the commodity index and the consumer price index as the main price factors for the forecast，combining with the quotient space theory and according to the time attribute，divided the gold price domain into three divisions of year，quarter and month，then established the three-tier quotient space，and conducted the granularity of the synthesis and calculation.The SVM forecasting model based on the quotient space theory was established to forecast the gold price.The forecast value of the gold price for year，quarter and month granularity is 8 122.4 CNY/troy ounce，7 947.506 CNY/troy ounce and 8 089.5 CNY/troy ounce respectively，and the composite result is 8 053.1 CNY/troy ounce.Finally，the reliability of the model is verified by comparing the gold price forecast with the GM（1，1） forecast of 9 382.2 CNY/troy ounce and the actual gold price of 8 306.0 CNY/troy ounce，indicate that the prediction result of the model is within the range allowed by the error，and the model is superior to traditional price prediction methods..

Keywords：gold price;influencing factor;quotient space;granularity;SVM model;Person correlation coefficient method;Granger causality test