Creating Predictive Rules

Understanding what drives an outcome within data is the end goal of predictive analytics. While most software attempts to create statistical models to fit new data, EmcienPatterns generates predictive rules that can be used to predict a specific outcome category in the data. Predictive rules can be seen as the top drivers and reasons for an outcome, and can be used by the software to make predictions on new data. In this article, we will use the laptop failure diagnostic log found in the Sample Data Set Repository.



To create predictive rules for an outcome, simply follow the steps in this example:

From the analysis screen, select the file you'd like to analyze along with the dependent category in the data that you'd like to create predictive rules for. 

During the analysis, the engine is going to use all of the connections it generates within the data set to find the data points that come together to allow the dependent category to be predicted. A rule is essentially a connection of one or more data points in the data set that give more information about the outcome than we would have without the rule.

After the analysis, click on the Rules link at the top of the dashboard to be taken to the predictive rules.

The engine has found all of the useful interactions within the data that lead to an outcome in the dependent category. Within the UI, these rules are grouped according to how often they happen:

  • The Strong Signals are the rules that have a 100% conditional probability to an outcome. Another way to think about these rules is that every time that collection of items was viewed within the data set, it had the same outcome
  • The Mixed Signals are rules with conditional probability of less than 100%, which give valuable insight and are used by the engine to make predictions 

The predictive rules can be thought of as drivers of the outcomes in the data set, and can be used to find the root cause of the different values in the dependent category of a file. To search for specific rules generated from the analysis, click on “Rules Search” on the right-hand side of the page, where you can filter and find rules for specific attributes.

After seeing the rules created during the analysis, you may want to filter to find specific rules containing certain values within your data. To search for specific rules generated from the analysis, click on “Rules Search” on the right-hand side of the page, where you can filter and find rules for specific attributes.

Although viewing the individual rules within the UI is helpful, the real benefit from creating predictive rules is using them to make predictions and automating the process.