Most Important Algorithm Parameters in C#

Printer Code 128 Code Set A in C# Most Important Algorithm Parameters

Higher support for a node means fewer splits and thus a shallower tree. A higher complexity penalty means fewer splits and thus a shallower tree. With the split method, you can force binary splits only or complete splits on all values of an attribute. This parameter specifies the maximum number of states allowed for an attribute. If you have more states, you have to discretize the values. This parameter influences the number of nodes in the hidden layer. If you set this parameter to zero, you have no hidden layer, so you create a Logistic Regression model. As the name of this parameter implies, you use it to control the number of clusters. This parameter specifies the minimum probability for a rule. You can have greater confidence in rules with higher probability.

This parameter specifies the minimum number of cases that must contain the itemset before a rule is generated. With lower minimum support and minimum probability, you get more rules. In this algorithm, the CLUSTER_COUNT parameter gives only an approximate number of clusters found. The algorithm adds additional clusters to this number if it finds distinguishable models. With this parameter, you inform the algorithm that you have periodical data. Periods are very typical for a time series. For example, if your time points are based on months, the same month repeats every twelfth time. With this parameter, you control the method used for training. Possible values are Auto-Regression Trees with Cross-Predict (ARTXP), ARIMA, or MIXED. The default is MIXED. This parameter specifies how the model should be mixed between ART and ARIMA algorithms to optimize forecasting. You can enter any value between 0 and 1 lower values favor ART and forecasting is optimized for fewer predictions, while higher values favor ARIMA and forecasting is optimized for more predictions.

You can set the algorithm parameters by right-clicking the mining model in the Data Mining Designer and then clicking the Set Algorithm Parameters option. Pr actice

In this practice, you will use advanced types of keys, including Key Sequence and Key Time columns, and nested tables to build association, sequence, and forecasting models. You will then create a clustering model, using a UDM cube as the source. You will also deploy the

models as a dimension of a cube. Last, you will modify the algorithms behavior by setting parameters and, in a clustering model, rename the clusters to meaningful names.

First, you need to build a model to explore association rules for model purchases. You will search for any sequences that exist in model purchases and then forecast quantities and amounts of sales of bike models in different regions.

1. 2. 3. 4. 5.

Open the TK 448 Mining Models Project you created in the previous lesson in BIDS. In the Solution Explorer window, right-click the Mining Structures folder, and then select New Mining Structure to start the Data Mining Wizard. On the Welcome page of the Data Mining Wizard, click Next. On the Select The Definition Method page, select From Existing Relational Database Or Data Warehouse. Click Next. On the Create The Data Mining Structure page, from the Which Data Mining Technique Do You Want To Use drop-down list, select the Microsoft Association Rules technique. Click Next. On the Select Data Source View page, verify that the Adventure Works DW2008 DSV is selected, and then click Next. On the Specify Table Types page, specify vAssocSeqOrders as the case and vAssocSeqLineItems as the nested table. Click Next. On the Specify The Training Data page, clear the check box showing that the CustomerKey column is the key column (it is selected by default), and specify OrderNumber as a key column and Model as a key, input, and predictable column, as Figure 9-9 shows. Click Next. On the Specify Columns Content And Data Type page, click Detect. Click Next. On the Create Testing Set page, set the percentage of data for testing to 0. Because this is not a predictive model, you do not need the testing set. Click Next. Type tK448 ch09 association as the mining structure name and tK448 ch09 association rules as the model name, and then click Finish In the Solution Explorer window, right-click the Mining Structures folder, and then select New Mining Structure to start the Data Mining Wizard. On the Welcome page of the Data Mining Wizard, click Next. On the Select The Definition Method page, select From Existing Relational Database Or Data Warehouse. Click Next. On the Create The Data Mining Structure page, from the Which Data Mining Technique Do You Want To Use drop-down list, select the Microsoft Sequence Clustering technique. Click Next. On the Select Data Source View page, verify that the Adventure Works DW2008 DSV is selected, and then click Next.

6. 7. 8.

9. 10. 11. 12. 13. 14. 15.