基于动态隶属度的模糊时间序列在我国居民消费水平预测上的应用

准确预测我国居民消费水平对促进经济持续协调发展具有重大的理论和现实意义。根据1952～2013年我国居民消费水平数据,本文提出了一种基于动态隶属度的模糊时间序列预测方法。首先对数据聚类并模糊化处理得到隶属度序列;然后再对隶属度序列进行时间序列分析建模得到预测值;紧接着对其去模糊化实现我国未来三年居民消费水平的预测;最后,将预测结果与传统时间序列方法预测结果相比较,新方法的预测平均绝对误差(MAE)、均方误根差(RMSE)较传统时间序列方法分别减少了23和28。结果表明,本文的预测方法相对于传统时间序列预测方法具有较高的预报精度,可用于居民消费水平的预测。

The Application of Fuzzy Time Series Based on Dynamic Membership Degree in Forecasting the Consumption Level of Residents in China

Accurately predicting the consumption level of residents in China has great theoretical and practical significance for promoting sustained and coordinated economic development. According to the data of Chinese residents’ consumption level from 1952 to 2013, this paper proposes a fuzzy time series prediction method based on dynamic membership degree. Firstly, the data is clustered and fuzzified to obtain the membership degree sequences. Then the time series analysis of the membership degree sequences is used to obtain the predicted values. And then the defuzzification of the predicted values is used to forecast the resident’s consumption level in the next three years in China. Finally, comparing the prediction results with the prediction results of the traditional time series method, we can know that the new method’s predicted mean absolute error(MAE) and mean square root error(RMSE) are reduced by 23 and 28 respectively, compared with the traditional time series method. The results show that the proposed method has higher forecast accuracy than the traditional time series method and can be used to forecast the consumption level of residents.