Managing Analysis Services Models

The MS Decision Tree modeling node is used in predictive modeling of both categorical and continuous attributes. For categorical attributes, the node makes predictions based on the relationships between input columns in a dataset. For example, in a scenario to predict which customers are likely to purchase a bicycle, if nine out of ten younger customers buy a bicycle but only two out of ten older customers do so, the node infers that age is a good predictor of bicycle purchase. The decision tree makes predictions based on this tendency toward a particular outcome. For continuous attributes, the algorithm uses linear regression to determine where a decision tree splits. If more than one column is set to predictable, or if the input data contains a nested table that is set to predictable, the node builds a separate decision tree for each predictable column.

The MS Clustering modeling node uses iterative techniques to group cases in a dataset into clusters that contain similar characteristics. These groupings are useful for exploring data, identifying anomalies in the data, and creating predictions. Clustering models identify relationships in a dataset that you might not logically derive through casual observation. For example, you can logically discern that people who commute to their jobs by bicycle do not typically live a long distance from where they work. The algorithm, however, can find other characteristics about bicycle commuters that are not as obvious. The clustering node differs from other data mining nodes in that no target field is specified. The clustering node trains the model strictly from the relationships that exist in the data and from the clusters that the node identifies.

The MS Association Rules modeling node is useful for recommendation engines. A recommendation engine recommends products to customers based on items they have already purchased or in which they have indicated an interest. Association models are built on datasets that contain identifiers both for individual cases and for the items that the cases contain. A group of items in a case is called an itemset. An association model is made up of a series of itemsets and the rules that describe how those items are grouped together within the cases. The rules that the algorithm identifies can be used to predict a customer's likely future purchases, based on the items that already exist in the customer's shopping cart.

The MS Naive Bayes modeling node calculates the conditional probability between target and predictor fields and assumes that the columns are independent. The model is termed naïve because it treats all proposed prediction variables as being independent of one another. This method is computationally less intense than other Analysis Services algorithms and therefore useful for quickly discovering relationships during the preliminary stages of modeling. You can use this node to do initial explorations of data and then apply the results to create additional models with other nodes that may take longer to compute but give more accurate results.

The MS Linear Regression modeling node is a variation of the Decision Trees node, where the MINIMUM_LEAF_CASES parameter is set to be greater than or equal to the total number of cases in the dataset that the node uses to train the mining model. With the parameter set in this way, the node will never create a split and therefore performs a linear regression.

The MS Neural Network modeling node is similar to the MS Decision Tree node in that the MS Neural Network node calculates probabilities for each possible state of the input attribute when given each state of the predictable attribute. You can later use these probabilities to predict an outcome of the predicted attribute, based on the input attributes.

The MS Logistic Regression modeling node is a variation of the MS Neural Network node, where the HIDDEN_NODE_RATIO parameter is set to 0. This setting creates a neural network model that does not contain a hidden layer and therefore is equivalent to logistic regression.

The MS Time Series modeling node provides regression algorithms that are optimized for the forecasting of continuous values, such as product sales, over time. Whereas other Microsoft algorithms, such as decision trees, require additional columns of new information as input to predict a trend, a time series model does not. A time series model can predict trends based only on the original dataset that is used to create the model. You can also add new data to the model when you make a prediction and automatically incorporate the new data in the trend analysis. See the topic MS Time Series Node for more information.

The MS Sequence Clustering modeling node identifies ordered sequences in data, and combines the results of this analysis with clustering techniques to generate clusters based on the sequences and other attributes. See the topic MS Sequence Clustering Node for more information.