Running the ML pipeline

Task Mining is integrated with UiPath® AI Center™ to run the machine learning model and analyze the dataset.

For a Data Analysis, the datasets you upload should contain at least 10,000 actions, but we recommend 50K actions as a good baseline. The model can support up to 200K actions collected in a project. As a quick way of estimating, an active user typically provides 200-1K actions per hour, depending on the scenario they are recording.

Note: Knowledge of data science techniques is not required because the model takes care of this for you.

To run the analysis, follow the steps below:

1. On the Projects page, select your project.

2. On the Overview tab you can see the percent and number of actions captured. If it meets the minimum requirements of 10,000 actions captured, the Run analysis button is available. Click Run analysis to start the process.

   
   

   

   
   

3. A pop-up window recommends having at least 50,000 actions captured for optimal results. If you don't reach the optimal number, you can click Run anyway.

   
   

   

   
   
   Important: Running an analysis consumes 5,000 AI Units per analysis, regardless of the number of actions. To find more information about AI Units, check out the AI Units documentation page.
   
   

   

   
   

4. Analysing the recorded actions can take a while, and an informational message is displayed saying that this action can take up to 6 hours to complete.

   
   

   

   
   

5. When the results of the analysis are ready, you'll receive an email. Inside the email, you can click See Discovered Tasks now to see the discovered tasks. The analysis results are uploaded automatically to the Task Mining project.

   
   

   

   
   

   If the amount of captured actions is insufficient for a good result, the analysis can be completed without significant results. You are informed about the results via email, and a banner is displayed on the Overview page. In this case, it's recommended to capture more actions.

Generating an analysis result might cause errors. If this scenario happens, a banner appears on the Overview page, with a clickable link to contact support to resolve the issue promptly.