Spatial data mining is the process of trying to find patterns in geographic data. Most commonly used in retail, it has grown out of the field of data mining, which initially focused on finding patterns in textual and numerical electronic information. Spatial data mining is considered a more complicated challenge than traditional mining because of the difficulties associated with analyzing objects with concrete existences in space and time.
As with standard data mining, spatial data mining is used primarily in the world of marketing and retail. It is a technique for making decisions about where to open what kind of store. It can help inform these decisions by processing pre-existing data about what factors motivate consumers to go to one place and not another.
Say that Ashley wants to open a nightclub on a certain city block. If she had access to the appropriate data, she could use spatial data mining to find out what spatial factors make night clubs successful. She might ask questions like: Will more people come to the club if public transit is nearby? What distance from other nightlife venues maximizes patronage? Is proximity to gas stations a plus or a minus?
Ashley might also want to ensure that the people who come to her nightclub arrive in an even distribution over the course of an individual night. She could also use spatial data mining—perhaps more accurately, spatiotemporal data mining—to find out how people move through the city at certain times. The same process could be applied to patronage on different nights of the week.
The difficulties of this method are a result of the complexity of the world beyond the internet. Whereas past efforts at data mining usually had databases ripe for analysis, the inputs available for spatial data mining are not grids of information but maps. These maps have different types of objects like roads, populations, businesses, and so on.
Determining whether something is "close to" something else goes from being a discrete to a continuous variable. This massively increases the complexity required for analysis. Incredibly, this is one of the more simple types of relationships available to someone attempting spatial data mining.
Spatial data mining also faces the problem of false positives. In the process of searching data looking for relationships, many apparent trends will emerge as a consequence of statistical false positives. This problem also exists for the task of mining a more simple database, but it is amplified by the magnitude of data available to the data miner. Ultimately, a trend identified by data mining should be confirmed through the process of explanation and additional research.
