Method and system for comparing populations of entities to make predictions about business locations
A method and system including software makes predictions about potential business locations. The method includes the steps of providing heterogeneous data including a spatial component, extracting entities from the heterogeneous data, clustering the entities, and making a prediction about the business location using the clustered entities. Entities preferably include stores, people, or other physical objects and each has attributes. Preferably the entities include at least one of the following attributes: Value Ratio, Focal Values, Impact, Revenue Difference, Support, and Baseline Value. In this way, predictions made about the business locations will account for the nature of the business and enable comparison of one business location to another.
1. A method for making and displaying predictions about a business location, comprising the steps of:
providing, from a data source, heterogeneous data including a spatial component and an interface resource, the interface resource being hardware, for enabling output of the predictions;
extracting entities from the heterogeneous data;
clustering the entities to form a population of entities; and
making and displaying a prediction about the business location using the population of entities.
2. A method as set forth in claim 1 , wherein the step of extracting includes profiling of the heterogeneous data to yield entities for the business location.
3. A method as set forth in claim 2 , wherein profiling of the heterogeneous data includes calculating entity attributes chosen from at least one of: Value Ratio, Focal Values, Impact, Revenue Difference, Support, and Baseline Value.
4. A method as set forth in claim 2 , further comprising: comparing the prediction about the business location with a prediction made for another business location, wherein the step of comparing directly compares the profiled entities of one location with the profiled entities made for another location.
5. A method as set forth in claim 1 , wherein the entities have attributes characterized by data points and the step of clustering employs a k-means algorithm.
6. A method as set forth in claim 5 , wherein the step of clustering includes: choosing the number of clusters k;
randomly generating k clusters and determining k cluster centers;
assigning each data point to the nearest cluster center;
recomputing new cluster centers; and
repeating the steps of determining and assigning.
7. A method as set forth in claim 5 , wherein the step of clustering data points includes:
choosing the number of clusters, k;
directly generating k random points as cluster centers;
assigning each data point to the nearest cluster center;
recomputing new cluster centers; and
repeating the steps of determining and assigning.
8. A method as set forth in claim 1 , wherein the step of clustering employs an expectation maximization algorithm.
9. A method as set forth in claim 1 , wherein the step of clustering employs spectral clustering.
10. A method as set forth in claim 1 , wherein the step of clustering employs principal components analysis.
11. A method as set forth in claim 1 , wherein the step of comparing compares a population of entities from the business location with a population of entities from another business location.