Based on the findings of data mining, marketers are able to propose a proper strategy to benefit from the strong association between diapers and beer. Photo credit: Growing Up Design
Do you know the relationship between diapers and beer? One supermarket chain in the United States organizes a sales promotion of diapers and beer every weekend and always enjoys profitable results. This is because the husband is ordered by his wife to pick up diapers for their baby on the weekend and he will never forget to buy beer for his own enjoyment. Through data mining, supermarket marketers are able to identify customer’s weekend behavior: a specific group of customers move to the liquor section for beer after they purchase diapers. With this discovery, the marketers are able to propose a proper strategy to benefit from the strong association of the two products.
Data mining is a software-based algorithm which analyzes all kinds of potential behavior and their implications and then identifies correlations or patterns in a large amount of data. These patterns can then be used for many applications.
In recent years, the study and applications of people's movement prediction through data mining are increasingly valued. Josh Jia-ching Ying, Postdoctoral Fellow of the National Chiao Tung University, says that there is only so much you can do if you only know about the present situation. However, if you know what may happen next, you can prepare in advance and a lot more can be done. For example, for marketing purposes, transportation management and movement management after a disaster, if you can predict the next movement of a specific group of people, you will be able to propose highly effective corresponding measures.
Vincent S. Tseng, Professor of Computer Science at the National Chiao Tung University says that GPS data of every mobile phone or vehicle can be recorded and uploaded to the cloud. The GPS data is in fact coordinates and serves as the source for data mining. If timestamp is added, a person’s coordinates at different times can be acquired.
The most important objectives of data mining applications are accuracy and speed. Past researches which only considered coordinates and time failed to incorporate geographic semantics such as financial districts, cultural areas or scenic spots, thus leaving much to be desired in accuracy. Professor Tseng is the first in Taiwan who considers all the three factors, coordinates, time and geographic semantics, and hence is able to reach high prediction accuracy.
For example, there two people who both enjoy going to cultural events on the weekend and one of them lives in Taipei and the other in Tainan. Their time and coordinate data may not have much of a relation but if the geographic semantics is added, we will know that they prefer going to cultural events in their spare time. Based on their similar behavior pattern, when the Tainan citizen visits Taipei, the Taipei citizen’s favorite spots can be recommended to the tourist, thus reaching better recommendation effect.
A tree structure is used to ensure a high searching speed when doing research. Ying adds that there are many behavior patterns and if they are saved one entry after another, it will take an enormous storage space. However, if relevant data is organized into a tree structure, the searching speed will be accelerated.
We often receive all sorts of event information in our mobile phones or emails and aimless marketing is prone to causing antipathy. If we can predict customers’ demands and send information at the right time, this kind of marketing is then considered successful. Thus it is a key to predict where the subject will move to next based on data mining. The current data mining has shifted its focus from frequent pattern to high utility. In addition to business, government services, data streams and biomedical science, data mining is also ideal for cross-disciplinary applications. That is to say, when prediction technology is fully developed, data mining will be applied to a broad range of areas.
Translated by Glen E. Lucas
Date:15 Nov 2016