Plot maps
A plot map can be used to accurately locate a set of records onto a map at their location and optionally colour them based on a particular attribute of the data. In the example below I’ve built on the multi-layer reference map above and selected all the properties sold in the suburb of Port Melbourne. Furthermore, I’ve then coloured the cyprus mobile number example plot points in this layer based upon whether they fit into my definition of low/medium and high priced properties. I also added a number of other data fields to the plot map so that hovering over a point would show relevant information about that property (8).
The Apteco Datathon: 4. The property market in Melbourne
In this section we’ll use some of the predictive modelling capabilities built into FastStats to try to explain what makes a property expensive. I’m going to use the top 10% percent of all property price sales (3,486) as my analysis set, and the whole set of properties (34,857) as my base set. We’ll begin by looking at a set of variables that we have available and see how accurate they are as predictors of whether a property is expensive or not.
In our dataset we’ve started with 22 possible predictors and my first step is to consider which variables that I might want to use within my model. One useful tool is the association matrix provided by the Modelling Environment tool, which allows us to see the correlations between our predictive variables. This information, in conjunction with my domain knowledge, allows me to select variables that I think will be important. It also allows me to remove variables that definitely aren’t going to be relevant or that are highly correlated with other variables.