Making Spatial Statistics Service Accessible On Cloud Platform

Xiangming Mu, Jing‐Hang Wu, T Li, Yanfei Zhong, Xiaorong Gao2014 · DOI

Abstract

Web services can bring together applications running on diverse platforms, allowing users to access and share data, information, and models more effectively from a centralized platform. Cloud computing has emerged as a paradigm of Internet computing providing dynamic, scalable, and virtualized resources as services. Addressing the calculation efficiency, maintenance costs, and data security issues of traditional platforms, this paper presents a spatial statistics service based on the Microsoft Azure cloud. Experimental results demonstrate that this service is accessible, convenient, and highly efficient for public use.

Introduction

Standard statistical methods, such as Ordinary Least Squares (OLS) regression, rely on the fundamental assumption that observations are independent. However, geographic data almost always violates this rule due to spatial autocorrelation—the tendency for neighboring locations to share similar characteristics. When researchers ignore these spatial patterns, their models become biased and statistically unreliable.

While specialized software like GeoDa exists to handle these issues, traditional desktop applications face significant hurdles in the era of "Big Data," including limited local processing power, high maintenance costs, and difficulties in sharing models across diverse platforms. This research bridges the gap by creating a "Spatial Statistics as a Service" (SSaaS) model that leverages the scalability of the cloud to provide robust spatial regression tools.

Methodology: Eigenvector-based Spatial Filtering (ESF)

The core of the system is the Eigenvector-based Spatial Filtering (ESF) method. This approach is particularly elegant because it "filters" out spatial autocorrelation, allowing the final analysis to utilize familiar OLS techniques while maintaining statistical integrity.

Flowchart detailing the iterative process of the Eigenvector-based Spatial Filtering algorithm.

As shown in the algorithm flowchart, the process begins by creating a spatial weight matrix and centralizing it. The system then calculates eigenvalues and eigenvectors, which represent various spatial patterns. Through an iterative process, the system selects eigenvectors that minimize Moran’s I (a measure of spatial autocorrelation) and tests them for significance. These selected eigenvectors are then added as independent variables to the regression model.

Cloud Implementation and System Architecture

The system was deployed on Microsoft Windows Azure, utilizing the .NET framework to manage web interfaces and parallel computing tasks. The backend mathematical functions were developed in C++ using Microsoft’s Component Object Model (COM) and compiled into Dynamic Link Libraries (DLLs) to ensure high-performance matrix operations.

The staService web interface (left) and the validation dataset of 30 Chinese provinces (right).

The user interface, titled staService, allows users to invoke various statistical operations—such as calculating Moran’s I, eigenvalues, and OLS regression—directly through a web browser. This removes the need for high-end local hardware or complex software installations.

Experimental Validation and Results

The authors validated the cloud service using a dataset of economic development indicators from 30 Chinese provinces. The study used Gross Domestic Product (GDP) as the dependent variable, with Land Area and Population as independent variables. A spatial connectivity matrix was constructed based on provincial borders.

The cloud-based ESF service significantly outperformed traditional OLS regression (tested via GeoDa) in terms of model accuracy and fit:

  • Model Fit (R-squared): The standard OLS model achieved an R-squared of 0.640, while the cloud-based ESF model reached 0.838. This indicates that the ESF model explained roughly 20% more of the variance in the data by accounting for spatial patterns.
  • Model Parsimony (AIC): The Akaike Information Criterion (AIC) dropped from 609.068 (GeoDa) to 580.476 (ESF). A lower AIC indicates a better-fitting model that avoids over-complexity.

Conclusion

This research successfully demonstrates a proof-of-concept for cloud-based spatial statistics. By transitioning complex spatial regression to a scalable API, the platform provides high processing efficiency and remote access "wherever and whenever" via a browser. Future work will focus on scaling the service for massive datasets (millions of points) and expanding the toolset to include more diverse spatial statistical methods beyond ESF.

Summary updated

Abstract

Web service can bring together applications running on diverse platforms, users can access and share various data, information and models more effectively and conveniently from certain web service platform. Cloud computing emerges as a paradigm of Internet computing in which dynamical, scalable and often virtualized resources are provided as services. With the rampant growth of massive data and restriction of net, traditional web services platforms have some prominent problems existing in development such as calculation efficiency, maintenance cost and data security. In this paper, we offer a spatial statistics service based on Microsoft cloud. An experiment was carried out to evaluate the availability and efficiency of this service. The results show that this spatial statistics service is accessible for the public conveniently with high processing efficiency.
BibTeX
@inproceedings{mu2014making,
  author = {Xiangming Mu and Jing‐Hang Wu and T Li and Yanfei Zhong and Xiaorong Gao},
  title = {{Making Spatial Statistics Service Accessible On Cloud Platform}},
  booktitle = {˜The},
  year = {2014},
  doi = {10.5194/isprsarchives-xl-6-73-2014},
}
Figure