CST(48)-Data Mining

Data-mining tools help you sift through vast quantities of information looking for valuable patterns in the data. A pattern may be as simple...

Data-mining tools help you sift through vast quantities of information looking for valuable patterns in the data. A pattern may be as simple as the realization that 80% of male diapher buyers also buy beer. Data mining is the process of discovering unsuspected patterns. Some early data-mining pioneers are reporting 1000% return on investment.

Data mining looks for patterns and groupings that you may have overlooked in your own analysis of a set of data. The tool typically performs a “fuzzy” search, with little or no guidance from an end user. In data mining, the tool does the discovery and tells you something, instead of you asking it a question. In contrast, query tools and OLAP return records that satisfy a query that you must first formulate. Instead of responding to low-level queries, data mining tools deal with fuzzy searches. They don’t assume that you know exactly what you’re seeking. Most tools use the following search methods:

•  Associations look for patterns where the presence of something implies the presence of something else. For example: “Scuba gear buyers are good candidates for Australian vacations.
•  Sequential patterns look for chronological occurrences. For example: the price of stock X goes up by 10%, the price of stock Y goes down by 15% a week later”.
•  Clustering’s look for groupings and high-level classifications. For example: “Over 70% of undecided voters have incomes of over Rs.60000, age brackets between 40 and 50, and live in XYZ neighborhood.”

The output of the discovery process is often represented in the form of if-then rules; for example:

     If
        Age = 42; and
        Car_Make = Volvo; and
        No_of_Children < 3
    Then
        Mailorder_Response = 15%