sluofoss

Recent Notes

6. Perform analytics in power bi

Jan 13, 20254 min read

Introduction to analytics

pbi make this more accessible to DA than traditionally DE.

WARNING

sections mentioned here but not in ms learn unit

  • Use advanced analytics custom visuals.
  • Review Quick insights.
  • Apply AI Insights.

Explore statistical summary

  • dax statistical functions like
    • sum
    • average
    • etc
    • better than implicit measure via the ui for performance
  • histogram + bell curve
    • binning
      • group type
        • number of bins
  • topn analysis
    • use qa visual to find topn
    • topn filter type
    • topn dax function

Identify outliers with Power BI visuals

  • use visual (scatter)
  • use dax
    • example here replicate normal result. not a statistical way to highlight outlier.

Group and bin data for analysis

  • group is for catrgories of data.
  • binning is for grouping continuous fields.
  • to create a group on the visual, ctrl+click to select data point > right click > select group data
  • Edit a group
    • Right-click the group field in either the Legend bucket or the Fields pane, and then select Edit Groups.
  • create bin groups
    • Start in the Fields pane by right-clicking the Order Qty field that you want to create the bins for.
    • then select New Group.
    • On the Groups window that displays, set the Bin size to the size that you want, adjust other settings as required, and then select OK.

Apply clustering techniques

in scatter visual > more options (… ) > automatically find clusters > edit fields, num of cluster.

  • applied clustering to the scatter allows only use two measures.
  • for clustering by using more than two measures, use a table visual instead,
    • add all the fields you want to use, and then run the clustering algorithm by using the same process.

Conduct time series analysis

  • appsource has animation custom visual called play axis (dynamic slicer).
    • act like dynamic slicer and display time trend without user interaction.
    • Play Axis visual that is available for scatter chart visualizations within Power BI Desktop has similar functionality
  • formatting of play axis
    • animation setting
      • auto start animation
      • looping
      • speed
    • colors
      • overall color and of each button
    • caption for visual.

Use the Analyze feature

NOTE

This feature does not work if you have non-numeric filters applied to your visual and/or if you have measure filters applied.

steps:

  • right-click a data point on the visual
  • hover over the Analyze option to display two further options:
    • Explain the increase
    • Find where the distribution is different.
  • The options that are available depend on the data point that you selected.

For more information about the Analyze feature, see Apply insights in Power BI Desktop to discover where distributions vary (preview).

Create what-if parameters

use slicer to control the parameter value and can be used to project growth and compare to previous performance.

Use specialized visuals

The three main AI visuals are:

  • Key influencers
    • find most correlated variable
    • Key influencers uses clustering and ML to determine relationships while decomposition tree already has the relationships determined through your schema
  • Decomposition tree
    • visualize data in multi dimension in any order
    • valuable tool for ad hoc exploration and conducting root cause analysis.
  • Q&A
decomposition treekey influencer
useful for root base analysisExplore the impact of a specific value on the change of a target variable
Composed of nodesDependent on enough observations to derive patterns
Allows the exploration of a target variable using different pathsUseful for directly comparing influence across variables

decomposition tree (datacamp)

powerbi decomposition tree (with AI split) can achieve what MDI/shapley with decision tree does.

it is similar to sankey but instead of count it shows a custom metric at each level.

relative vs absolute ai split

  • absolute (find variable with greatest max min difference for all vars)
  • relative (find variable with single category of highest deviation within this var)

https://anthonysmoak.com/2023/11/27/the-power-bi-decomposition-tree-guide-for-data-analysis/#:~:text=Another%20benefit%20of%20the%20decomposition,the%20measure%20you%20are%20analyzing.

some semantics and keywords for Q&A visual (datacamp)

x by y group by z as plot type

Graph View

Backlinks