These tools are the keys to the analysis of data in the data warehouse, enabling corporate business analysts to effectively 'mine' customer data:
Neural Networks
A neural network consists of three interconnected layers: an input and output layer with a hidden layer between The hidden processing layer is like the brain of the neural network because it stores or learns rules about input patterns and then produces a known set of outputs. Because the process of neural networks is not transparent, it leaves the user without a clear interpretation of the resulting model.
Decision Trees
Decision trees divide data into groups based on the values of different variables. The result is often a complex hierarchy of classifying data, which enables the user to deduce possible future behavior. For instance, it might be deduced that for a person who only uses a credit card occasionally, there is a 20% probability that an offer for another credit card would be accepted. Although decision trees usually are faster than neural networks, they have drawbacks. One of these is the handling of data ranges as in age groups, which can inadvertently hide patterns.
Rule Induction
The method of rule induction is applied by creating non-hierarchical sets of possibly overlapping conditions. This is accomplished by first generating partial decision trees. Statistical techniques are then used to determine which decision trees to apply to the input data. This method is especially useful in cases where there are long and complex condition lists.
Data Visualization
Data visualization is not really a data mining tool, however, because it provides a picture for the user with a large number of graphically represented variables, it is a powerful tool for providing concise information. The graphics products available for data visualization make the detection of patterns much easier than when more numbers are analyzed.
Because of the pros and cons of the varied data mining tools described above, software vendors today incorporate all or some of them in their data mining software packages. Each tool is essentially a matter of looking at data with different means and from different angles.
Analysis and Interpretation
The last step in the data mining process is analyzing and interpreting results. The extracted and transformed data is analyzed with respect to the user’s goal, and the best information is identified and presented to the decision-maker through the decision support system. The purpose of result interpretation is not to only represent the output of the data mining operation graphically, but also to filter the information that will be presented through the decision support system. For example, if the goal is to develop a classification model, during the result interpretation step, the robustness of the extracted model is tested using one of the established methods.
If the interpreted results are not satisfactory, it may be necessary to repeat the data mining step, or to repeat other steps. This situation relates directly to the quality of data. The information extracted through data mining must be ultimately comprehensible. For example, it may be necessary, after interpreting the results of a data mining operation, to go back and add data to the selection process or to perform a different calculation during the transformation step.
 |
