Lesson 1: Preparing and Creating Data Mining Structures in C#.NET

Creating Code 128 in C#.NET Lesson 1: Preparing and Creating Data Mining Structures

figure 9-8 The Dependency Network viewer, showing the three variables with the most influence on the Bike Buyer variable

8. 9.

In the Attribute Discrimination viewer, you can see the values of input attributes that favor value 1 of the Bike Buyer attribute and the values that favor value 0. Check the Clustering model as well. The viewers for the Clustering algorithm are similar to those for Na ve Bayes, although the Cluster Diagram view, which shows the number of clusters and how close these clusters are to one another, is different. You should see 10 clusters. You can shade the clusters based on a value of any variable. For example, you can use the Bike Buyer as the shading variable and select state 1 of the Bike Buyer variable to quickly see the clusters that have more bike buyers. In the Cluster Profiles viewer, you can see the distribution of values of variables in each cluster. In the Cluster Characteristics viewer, you can see the values that are characteristic of a specific cluster. Using the Cluster Discrimination tab, you can compare pairs of clusters to identify the values of input variables that favor particular clusters of the pair. Check the Neural Network model. The only viewer you will see is the Attribute Discrimination viewer, in which you can identify the values of input attributes that favor value 1 of the Bike Buyer attribute and values that favor value 0. In addition, you can filter the viewer to show the discrimination for specific states of input attributes only. Last, also check whether you got a different tree for the filtered Decision Trees model compared to the unfiltered one. Save the solution, and then close BIDS.

11. 12.

2. What are the different components that make up the SQL Server BI suite 3. Why would you prepare separate training and test sets 4. Which algorithm is appropriate for fraud detection 5. In the practice, why do you think you had to discretize the continuous attributes 6. Do you get the same results when you use different algorithms for predictions 7. How can you prepare training and test sets

mining. you should use a predictive algorithm such as Decision Trees to discover patterns in the data that you have about customers who have already left your company. you should then use the same algorithm to predict the churn based on data from existing customers.

2. The SQL Server BI suite includes SSAS cubes, SSAS data mining, SSRS, and SSIS. 3. Whenever you use predictive models, you have to test the predictions. you train

the models by using the training sets and then test the models by making predictions with the trained models on the test set.

5. you had to discretize the continuous attributes because of the na ve Bayes algorithm, which accepts discrete attributes only. If you did not discretize them, you would have to ignore them in the na ve Bayes model.

give you different results for example, slightly different predictions. That is why you have to evaluate the models by using them to make predictions on the test set.

Mining Designer in BIDS to specify the percentage of the holdout data for the test set. you can also use the TABLESAMPLE option of the T-SQL SELECT statement or the SSIS Row Sampling Transformation and Percentage Sampling Transformation.

In Lesson 1, you used a relational source to create quite a few predictable models, and you used columns from a relational source for different purposes in your mining structure and models. In this lesson, you will learn more about column usage and about an advanced column type called a nested table. You will then use this knowledge in a practice to create association, sequence, and forecasting models. You will also use a cube as the source for a data mining model and learn how to set up the most important parameters for different data mining algorithms.