Taxonomy-based system for discovering and annotating geofences from geo-referenced data

1. A computer executable method for discovering functional clusters in an area of interest, comprising:

dividing a geographical area of interest into a plurality of cells representative of the geographical area of interest, wherein a cell includes a plurality of geo-referenced points, and wherein each geo-referenced point is associated with a label from among a plurality of labels;

assigning each cell an initial label corresponding to a label associated with a geo-referenced point included in a respective cell;

applying hierarchical clustering to create clusters by assigning two or more adjacent cells to a cluster based on a maximization of an objective function that favors spatially adjacent cells sharing a common label;

limiting overgeneralization of the common label while applying the hierarchical clustering;

updating boundaries on a map of the geographical area of interest to identify boundaries of the clusters; and

upon detecting a user moving from a first cluster of the clusters to a second cluster of the cluster, notifying a device associated with the user that the user is crossing from the first cluster to the second cluster, wherein the first cluster represents a first portion of the geographical area of interest and the second cluster represents a second portion of the geographical area of interest,

wherein the hierarchical clustering comprises:

assigning each cell to a new cluster, such that a number of clusters equals a number of cells;

computing, for each cluster, a contribution to the objective function for each label contained within the cluster to find a label having a maximum value for the objective function for the cluster;

assigning, for each cluster, a cluster label identifying the label having the maximum value for the objective function;

identifying a cluster pair of at least two clusters spatially adjacent to one another;

for the cluster pair, computing a contribution to the objective function for each cell within the cluster pair using the initial label assigned to the cell, and calculating a cluster pair contribution by summing each contribution to the objective function for each cell within the cluster pair;

for the cluster pair, calculating, for each possible label, a total contribution to the objective function for the cells contained in the cluster pair to find a maximum merged contribution;

for the cluster pair, comparing the cluster pair contribution to the maximum merged contribution to find a higher contribution;

for the cluster pair, when the maximum merged contribution is higher than the cluster pair contribution, assigning the cluster pair to a priority queue ranked according to the maximum merged contribution of the cluster pair; and

for the cluster pair in the priority queue:

merging a top ranked cluster pair in the priority queue to have at least one common label:

removing the merged top ranked cluster pair from the priority queue;

updating the priority queue by replacing references to clusters in the merged top ranked cluster pair with a reference to the merged top ranked cluster pair for each cluster pairs remaining in the priority queue; and

updating contributions of each cluster remaining in the priority queue.

2. The computer executable method of claim 1 , wherein the initial label is a most popular label within the cell.

3. The computer executable method of claim 1 , wherein the objective function is defined as:

Σ a i ∈C k (λΣ j∈[1,n] w i,j ( l *( a i )= l *( a j ))+(1−λ)(cov( l *( a i ), a i )−cov( l *( a i ), A )));

wherein:

a i is a cell;

C k is a cluster;

l is a cell label;

λ is a user-defined parameter between 0 and 1; and

cov(l(a i ) is a coverage function.

4. The computer executable method of claim 1 , wherein assigning each cell an initial label comprises assigning each cell an initial label based upon a type of commercial activity that occurs with the geographical area of interest represented by the cell.