Discovering Features Using Analytics and AI Visuals

In Chapter 4, you discovered features within the FAA Wildlife Strike data to be tested with Power BI ML to see whether they are good predictive features, built out queries in Power Query to structure those features into tables, and completed the foundation of your high-level data architecture. You are ready to take a deep dive into analytics and prep for ML in Power BI.

In this chapter, you will continue to build out your Power BI analytical reports as you explore data and discover trends. You will also try a few of the Power BI AI visuals to see if you can uncover additional features that could be added to your machine learning efforts. At the end of this chapter, you will have a more complete analytical report and more robust data to build, train, and test your Power BI ML models.

Technical requirements

For this chapter, you’ll need the following:

• Power BI Desktop April 2023 or later (no licenses required)
• FAA Wildlife Strike data files from either the FAA website or the Packt GitHub site at this link: https://github.com/PacktPublishing/Unleashing-Your-Data-with-Power-BI-Machine-Learning-and-OpenAI

Identifying features in Power BI using a report

Now that you’ve built out the base queries for Predict Damage, Predict Size, and Predict Height, you can add additional features to evaluate in each of the queries that will be used for Power BI ML. Features to evaluate the ML models will be added to each ML query. Building the ML models in Power BI will allow you to narrow down the list of features to those that are most useful. In order to make the chapter easy to read, you can take a look at potential new features one by one as they pertain to the three queries for Power BI ML. When training and testing your ML models in Power BI, the predictive value of each feature will be evaluated when you build the ML models. Therefore, if a feature is in question, you’re better off keeping it in the query and then removing it later on if it turns out to provide little value.

Note that if you don’t want to build all of these Power BI report pages as part of your journey...

Identifying additional features using the key influencers visual in Power BI

Power BI has a built-in AI visual called key influencers that can be used to discover interesting patterns in data. You will now take a look at some additional columns from the FAA Wildlife Strikes data and explore how they influence the likelihood of damage and the size of the animals.

Start by duplicating the Predict Damage page and renaming the copied version to Predict Damage Key Influencers. By adding a duplicate page, you preserve the filters in the Filter panel. Delete all of the visuals on the page. In the Visualizations pane, add the Key influencers visual to the page. Now follow these steps:

1. With the Key influencers visual highlighted, add the Indicated Damage field to Analyze.
2. Add the Phase of Flight, Precipitation, Sky, and Effect on Flight fields to Explain by.
3. Set the selection at the top of the visual for What influences Indicated Damage to True.

You notice that many...

Adding new features to the ML queries in Power Query

Back in Power Query, you will now add the new features that were discussed in this chapter to the ML queries. Looking through the list, Month Num is the only one that will need some custom M code to convert Incident Date to a month. You’ve already done this in the previous chapter, so you can reuse the Date.Month([Incident Date]) M code for a custom column.

Starting with Predict Damage highlighted in the group ML queries, follow these steps:

1. Double-click on Remove Other Columns under Applied Steps.
2. Add each of the features in the table shown in Figure 5.16 (including Incident Date to be converted into Month Num).

Your screen should look like this:

Figure 5.16 – Select the columns to be added to the ML query

3. Click OK.
4. Phase of Flight, Sky, Precipitation, and Effect on Flight all have some empty values. Replace the null (empty) values with the text value empty...

Summary

In this chapter, you discovered new features and added them to pages of your Power BI report. You also used the Power BI key influencers visual to explore the FAA Wildlife Strike data to find interesting correlations. You then added the new features to your Predict Damage, Predict Size, and Predict Height ML queries. Finally, you ensured that the ML queries were cleaned up and ready for Power BI ML.

In the next chapter, you will continue to explore the FAA Wildlife Strike data using different types of capabilities in Power BI such as R and Python visuals. Newly discovered features will then be added to your ML queries as you finish them up and prepare to graduate to the Power BI cloud service.