Distributed Intelligence for Constructing Economic Models

Distributed Intelligence for Constructing Economic Models

Ting Yu (University of Sydney, Australia)
DOI: 10.4018/joci.2010070102
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

This paper presents an integrated and distributed intelligent system being capable of automatically estimating and updating large-size economic models. The input-output model of economics uses a matrix representation of a nation’s (or a region’s) economy to predict the effect of changes in one industry on others and by consumers, government, and foreign suppliers on the economy (Miller & Blair, 1985). To construct the model reflecting the underlying industry structure faithfully, multiple sources of data are collected and integrated together. The system in this paper facilitates this estimation process by integrating a series of components with the purposes of data retrieval, data integration, machine learning, and quality checking. More importantly, the complexity of national economy leads to extremely large-size models to represent every detail of an economy, which requires the system to have the capacity for processing large amounts of data. This paper demonstrates that the major bottleneck is the memory allocation, and to include more memory, the machine learning component is built on a distributed platform and constructs the matrix by analyzing historical and spatial data simultaneously. This system is the first distributed matrix estimation package for such a large-size economic matrix.
Article Preview

Introduction

The input-output model of economics uses a matrix representation of a nation’s (or a region’s) economy to predict the effect of changes in one industry on others and by consumers, government, and foreign suppliers on the economy (Miller & Blair, 1985). Because the economic constantly evolves, the input-output model needs to be updated at least annually to reflect the new circumstance. Unfortunately, in most countries such as Australia, the input-output model is only constructed every 3-4 years, because the large amount of monetary and human cost is involved to complete a survey (ABS, 2007a). The Centre for Integrated Sustainability Analysis (ISA), University of Sydney, is developing an integrated intelligent system to estimate and update the input-output model at different level on a regular basis.

The input-output model consists of a time series of matrices (Table 1) representing the industry structure of a given year. At Table 1, each entry represents the commodity flow between the industry sections of different regions. For example, the entry represents the commodity flow from the sheep industry at Victoria (VIC), a state of Australia, to the retail industry in China. In this example, the goods worth of 0.23 million dollar are sold. Often is missing as it is very detailed information and hard to be surveyed. However government agents published aggregated information more frequently, such as V and U which represent the total export of a given industry in Australia (ABS, 2007b). The aggregated information is available for a rather long period, for example agriculture information from 1861 to 2007 in the database (ABS, 2007b). It is worth clarifying that the aggregated information is not limited by the sums of rows or columns as the V and U. The main purpose of this distribute intelligent system is to utilize those available aggregated information and the economic models from previous years to populate and update current or future s to build a series of this economic models for current year or coming years.

Table 1.
An example of the input-output table defined by the 3-level tree and the 2-level tree
China (1)U
Shoe (1)Retail (2)
Australia (1)NSW (1)Sheep (1)
Oil (2)
VIC (2)Sheep (1)
Oil (2)
V

Complete Article List

Search this Journal:
Reset
Open Access Articles
Volume 7: 4 Issues (2017)
Volume 6: 4 Issues (2016)
Volume 5: 4 Issues (2015)
Volume 4: 4 Issues (2014)
Volume 3: 4 Issues (2012)
Volume 2: 4 Issues (2011)
Volume 1: 4 Issues (2010)
View Complete Journal Contents Listing