Group 3 Overview

PROPOSAL

POSTER

APPLICATION

REPORT

Abstraction

Geospatial analysis was developed for problems in the environmental and life sciences, which has currently extended to almost all industries including economy, defence, utilities, social sciences and public safety. The application of geo-visualization using geographically weighted regression (GWR) is an exploratory technique mainly intended to indicate where non-stationarity is taking place on the map. It is a good exploratory analytical tool which creates a set of location based parameter estimates, able to be mapped and analysed to give spatial information for the relationship of explanatory variables and response variable.

Our study uses economy data to explore district GDP condition in east region of China. The project scope covers the analysis, model and visual representation of multivariate factors like GDP,Industry Output, Usual Residence,Average Wage,Area,City Construction Rate,No. of higher institution, and ratio of Teacher/Student which contributes to economical development in each city area of the province or municipality with the assistance of interactive charts and graphs.

Conventional linear regression models are commonly used to analyse environmental problems to see the influences and basic relationship of factors which contributes to the economy. We feel that a geographical regression model can also be built to make spatial statistical analysis for the issues related to economy. GWR is a technique by allowing local instead of global parameters to be estimated. The model fits where a localized adjustment conveys a more meaningful message with involvement of spatial areas. It uses a moving window weighting mechanism for localised models detected at target location. Results are mapped to an interactive exploratory geo-application with the consideration of the nature of data spatial heterogeneity.

Motivation

Conventional linear regression models are commonly used to analyse economical problems to see the influences and basic relationship of factors which contributes to the economy. We feel that a geographical regression model can also be built to make spatial statistical analysis for the issues related to economy. GWR is a technique by allowing local instead of global parameters to be estimated. The model fits where a localized adjustment conveys a more meaningful message with involvement of spatial areas. It uses a moving window weighting mechanism for localised models detected at target location. Results are mapped to an interactive exploratory geo-application with the consideration of the nature of data spatial heterogeneity.

Objective

Our objective is to provide an interactive and exploratory geo-visual analytics tool for regional and urban planner and policy maker to visualize, analyse and model the location-based data.

Data Source and Preparation

The following data sources will be used for the project:
1. Economy data from National Bureau of Statistics of China in 2015.

Variable ,Full Name ,Unit
City ,City Name ,-
Province_Municipality ,Province or Municipality Name ,-
GDP_billion_b ,GDP Value ,RMB (billion)
Primary_Industry_b ,Primary Industry Output Value ,RMB (billion)
Secondary_Industry_b ,Secondary Industry Output Value ,RMB (billion)
Teriary_Industry_b ,Teriary Industry Output Value ,RMB (billion)
Usual_Residence_k ,No. of usual residence ,thousand
Average_wage_RMB ,Average wage ,RMB (digit)
Area_sqkm ,Total Area Size ,sqkm
City_Construction_Rate ,Rate of city construction ,%
No_Higher_Institution ,No. of higher institution ,digit
Teacher/Student ,Ratio of teacher vs student ,%

Approach

The main framework used for the application is RShiny and Leaflets. There are some packages in R to build a smooth interactive interface for extracting, exploring, and modelling the data. We implemented GWR model and parallel coordinate graph for spatial analysis, which enables us to study the impact and influence of various factors contributes to the local economical growth situation. The interactive graphs and charts has also been created to allow the user to apply different scenarios and make deep exploration.

Selection of Tools

Since this course aims at pursuing the good command of R language, for the display part we will choose R to implement our visualization. To be more precise, we will use R Markdown and the other relevant package like spgwr, pryr, gwmodel, tidyverse, rgdal, RColorBrewer, lubridate, ggplot2,ggraph, and JavaScript library like d3.js, leftlet.js. At the data preparation step, we will take advantage of JMP Pro and Tableau as well.

Challenge

1.The preparation of dataset to fulfil the minimum model requirement
2.The interaction of R graphs and maps with Shiny.
3.The building of an effective and accurate GRW model in R.
4.Find the most suitable tools and libraries to implement the visual features.

Reference

[1] https://rstudio-pubs-static.s3.amazonaws.com/44975_0342ec49f925426fa16ebcdc28210118.html
[2] https://www.rdocumentation.org/packages/spgwr/versions/0.6-32/topics/gwr
[3] https://cran.r-project.org/web/packages/spgwr/vignettes/GWR.pdf
[4] http://csiss.org/GISPopSci/workshops/2011/PSU/readings/Grose-Brunsdon-Harris-GWR.pdf