Configuring the HANA Optimization Function

As an expert user, you can create a HANA Optimization Function. The function can be thought of as a powerful calculator that enables you to solve complex optimization functions. Using the functionality, you create an objective function with linear constraints to calculate how best to optimize an aspect of your business. An optimization example is to maximize profits on a product or store location. You can also save the optimization function for future use. Currently, the function is unavailable for use in predictive modelling chains.

To configure the HANA Optimization Function, ensure first that you are connected to an SAP HANA data source.
  1. In the Predict room, you open a new optimization function in the list of components in the right-hand panel by selecting add_component.gif Optimization Function.
    Note To edit an existing function, go to Predict - Optimizations - + (plus sign) - Functions - [Function Name].
  2. On the Objective Function tabbed page of the HANA Optimization Function, enter the following information:
    1. Enter a unique Function Name.
    2. Optionally, enter a Function Description.
    3. Select a Function Optimization Type, which can be Maximize or Minimize. For example, your function might be used to maximize profits or minimize costs.
    4. In the Function text box, enter an objective, linear optimization function in the format, 5x + 8y + (-4)z. Note that you must include plus (+) signs as separators between monomes (e.g. 5x). Monomes consist of coefficients and variables.
      Note To maximize the screen, click the Expand icon.
    5. Click Validate Function. The validation ensures that your function is a mathematical expression that is supported by the underlying Optimization Function Library (OFL). If the function passes validation, you are allowed to configure its variables. If not reenter the function following the formatting rules outlined in the previous step.
    6. Configure the optimization function variables. Choose a variable type of Nominal (integers), Continuous (the resulting values for the variable do not need to be integers) or Binary (one or zero); Select a positive or negative range (+/-), or both; optionally, enter an alias that more accurately identifies your variable, such as Product or Store Location.
      Note Optionally, you can click Save to store the function in the Optimization section of the component list in the Predict room.
    7. Click Next to move to the next screen.
  3. On the Function Constraints tabbed page, enter the following information:
    1. In the Function Constraints text box, enter any amount of linear constraints. The constraints are enforced to maximise (or minimize) your objective functions.
      Note Enter linear constraints in the format, 5x + 8y >= 400; 16y +(-4)z >= 200. Note that you must include semicolon (;) signs as separators between constraints, except for the final constraint. Each constraints must include monomes (e.g. 5x+8y), a constraint type (e.g. >=) and a constraint value (e.g. 400). Otherwise, the application returns an error.
    2. Click Validate Constraints. If your constraint(s) throw an error, consult the rules outlined in the previous step and re-enter your constraints.
    3. Click Solve.
      Note Optionally, you can click Save to store the function in the Optimization section of the component list in the Predict room. You can edit the function to try when you reopen.
  4. Click Results to open the Results page.

In the Results page, you view the optimization result. For example, consider a result of x:130.0 and y:20.0. In this case, to maximize or minimize your function, the output for x (or its alias) should be 130.0. And for y (or its alias), the output should be 20.0. The results also display an overview of the function and constraints that you entered.

Optionally, you can do the following:
  • Select Split Lines checkbox to view each function on a separate line for easier readability.

  • Select the Use Alias checkbox to list your function variables by their chosen aliases.

  • Copy and paste the results using the standard short cut keys, Ctrl A and Ctrl C, into a text editor or report tool.

  • Click Save to store the function in the Optimization section of the component list in the Predict room.

You can now configure the HANA Sentiment Analysis component and use it as a preprocessing step in a complex analysis.