Preparing required data for problem code recommendations

Before you can enable the AI configuration for problem code recommendations, you must prepare your training and inference data, review your training and inference filters, and manually select the work orders to include in training.

1. Review the training and inference filters.

   Filters determine what work orders can be used for training, retrieve all of the work orders that are used for inferencing, and determine what work orders have the problem code recommendation feature enabled. The filters are defined as query definitions for the MXAPIWODETAIL object structure. For problem code recommendations, the object structure and the query definitions are available by default, but you can edit the query definition to change what work orders can be included in training or inferencing.

   1. In Maximo Manage, in the Object structures application, open the MXAPIWODETAIL object structure.
   2. In the More actions menu, click Query definition.
   3. In the Queries to be assigned table, click the Filter icon and locate the AITRAINFILTER query, which is used for training, and the AIINFERENCEFILTER filter, which is used for inferencing.
      Select a filter and then in the Query Clause field, review the WHERE clause and change as needed for your use case. For the AITRAINFILTER filter, if ai_usefortraining=1 is removed from the Query Clause field, you can manually select work orders for inclusion in training but the work orders are not added to the training data.
      The following text is an example of the training filter.

      worktype in ('EM', 'CM') and ai_usefortraining = 1

      The following text is an example of the inference filter. You do not manually select work orders for the inference filter. The filter retrieves all qualifying work orders. For a work order to have problem code recommendations enabled, the work order must be included in the inference filter.

      worktype in ('EM', 'CM') and status in (select value from synonymdomain where domainid ='WOSTATUS' and maxvalue in ('WAPPR')) and reportdate > CURRENT DATE - 30 DAYS and ai_usefortraining = 0

   4. Click OK.
2. Select work orders for training.

   From the work orders that are included in the training filter, you must select at least 20 work orders for training. 10 of those work orders must contain one problem code. The other 10 work orders must contain another problem code. To decide what work orders to choose, consider the following best practices:

   • Use unique work orders. The training process filters out duplicate work orders.
   • Use diverse work orders. Ideally, work orders include a range of descriptions and address a range of problems.
   • Use work orders that have accurate problem codes.
   • Ensure that each problem code has an accurate description. You can edit problem code descriptions in the Failure Codes application.
   • Ensure that each work order has an accurate description. You can edit work order descriptions in either the Work orders or Work Order Tracking applications.

   • Although the minimum is 20 work orders, the larger, more diverse, and accurate your training filter data is, the more likely the model can accurately recommend problem codes. The ideal training filter contains 20 - 50 work orders per problem code.

   1. In Maximo Manage, open Work orders.
   2. Select work orders to include in the training filter. Ensure that the work orders qualify for the training filter that you reviewed in step 1.

      To select individual work orders, open a work order and then in the Actions menu, click Add to AI training model. A tag appears in the work order to indicate that the work order is added to the filter. Ensure that you save the work order.

      To select work orders in bulk, in the Work orders table, first ensure that the Problem code column is added. To add the column, click Manage columns and then select the column in the list. After the column is added, click the Filter icon and then filter the work orders by the problem code that you want to include in training. Select the work order check boxes and then click Add to AI training model.

      Note: The Add to AI training model action is applicable to only the training filter for only problem code recommendations. The action is available for all work orders, but if the query clause for the training filter does not include a work order, selecting the action for that work order causes the tag to appear on the work order but does not add the work order to the training data. If a work order qualified for the filter and was selected but then later was disqualified from the filter, the tag remains on the work order.