Application Of Prediction In Capita Consumption Expenditure Based On Principal Component Analysis And Neural Network

With the rapid and steady development of China’s economy, consumer demand in driving the country’s economic growth has played an increasing role in terms of country’s economy, particularly, under the slow growth of the world economy in recent years, the conditions of foreign trade in the national economy is gradually decreased, and consumption spending has become the main driving force of our economy, our country in the development of policies to promote consumption aspect even more attention, therefore, research spending characteristics of our residents can provide the correct reference for national policy and regional development, study of income and expenditure changes that can develop better policies to promote consumption of countries and regions.The main content of this paper is to analyze the residents consumption of country’s 31 provinces, municipalities and autonomous regions. Data source of this article is "China Statistical Yearbook-2015" in the statistical data related to the consumption of the relevant sub-regional residents.This paper mainly completed the following tasks.By using a neural network algorithm we predict capita consumption expenditure of 31 provinces, municipalities and autonomous regions.Before neural network algorithm, we do principal component analysis for the data. After data reduction, then we use neural network to predict.Principal component analysis is also known as the principal component method.Principal component analysis algorithm is a statistical algorithm, and principal component analysis algorithm to index the raw data variable dimensionality reduction,which reduced the high dimensional information low-dimensional information,simplifying data structure, while the principal component analysis algorithm is also capable of data streamline using a few main components to a surrogate variable of the original data.Neural Network was based on the idea of neuronal cells, and neurons are the basic processing unit of the neural network. Neural networks utilizing its unique characteristics to achieve the purpose of study and training, looking for potential law and relational data.Through data mining algorithm of principal component analysis and neural network algorithm, the main component of the neural network, which is the comprehensive utilization of the advantages of principal component analysis and neural network algorithm, reduce the dimension simplify the data using principal component analysis, then it can get the main components which can be the main component of the neural network’s input, so it could get better prediction.We do research on relationship between related indicators of this year’s resident consumption and the coming year’s capita consumption expenditures.In "China Statistical Yearbook-2015", the statistical data which related to the consumption of the relevant sub-regional residents, mainly comprised eight expenditure data of rural residents and urban residents, so the total number is sixteen indicators, and based on 16 year of spending predictors of total consumer, it indicated the spending in the coming year.In this article, we use three prediction algorithms to achieve the forecast. These three methods are the multiple linear regression algorithm, principal component regression algorithm and RBF neural network algorithm.Data from 2001 to 2012 was sample data to build predictive model of the sample,and using the sample data in 2013 as a test sample, it predicted the consumption expenditure in 2014 in 31 provinces, municipalities and autonomous regions.By linear regression algorithm, principal component regression algorithm and RBF neural network algorithm in this paper, it achieved the different forecasts, and several predicted results were analyzed, it found that RBF neural network algorithm effected well.