Techniques used in Data Mining

If you find this page all a bit too much for you, don’t worry! We’re here to take the strain.

These are the Data Mining techniques used to acquire that secret advantage we spoke of.

Decision Trees/Rules

We uses these when estimating difficult and expensive directions usually encountered in the financial campaigns. For example, one key decision was to decide the risk/payback of several thousands of members pounds to get a legal opinion that might or might not change a course of strategy. The decision tree recommended to spend the money. Another was to decide whether a claim should be pursued via the Financial Services Compensation Scheme, a solicitor, or both at the same time but pay more fees on a win. The clear answer was to do both and pay the extra.

Regression

This really means making sense of large tables, identifying what the bigger issues are to help focus, looking for patterns and correlations. For example, one company with an uneven web sales pattern over several years were at a loss to know why it varies. We found their sales were heavily dependent on the Consumer Confidence Index which was going through large gyrations at the time.

Clustering

This we use for identifying how market sectors sub divide, and the three or four main characteristics of each dominant subsector. We use this in conjunction with Association Rules.

We have also used Clustering to unmask an American climate denier writing an Australian blog and pretending to be living in the country.

Association Rules

This is a way of seeing into a large body of data or text, and determining what two or three topics seem to go together. A common example is a supermarket online store finds people might tend to buy cookies, milk, and eggs. A customer places an order for milk and eggs, and the system flashes up “Did you forget the cookies?” This might be a surprise for this particular customer, but the Association Rules suggested it was worth reminding them. This is hard to see in large shopping cart lists, but made simple through Data Mining.

Text Mining

This is proving an absolute diamond cutter in knowledge research. Through this, we are able to find out what your customer is thinking, the order they are thinking about it, and the language used in their minds. When used with Rules of Association, such a clear picture comes out that when we write an email series based on the research, we get much more reaction to the emails, and the followup teleselling appointment setting is notably more effective.

Statistics (descriptive)

These are the better known sets of statistics that tell you how skewed and in which direction your shopping cart sales might be. A notable finding in one case was a client who had set a minimum order value of £45 had a large spike in low number orders just over it, then a big gap between that and £100. Clearly customers had wanted something, saw the restriction, added a bit more to get over the limit, and bought. However it brought up the realisation that they had no idea how much business just under £40 was going to a competitor without the restriction!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s