Cross-filtering performance is too slow

[Jackson123]

hi,there, 

At present, the solution that encounters a PBI problem can achieve the requirements but the performance is slow, so I would like to consult whether there is a fast implementation solution.

Label the machine codes in categories, and then filter the results of the crossover section, and the labeled classification will have crossed machine codes, that is, the same machine code may belong to either category A or category B.

The current scheme is to generate two identical tables, filtered by machine code bidirectional many-to-many relationship. The effect can be achieved, but it is too slow, the actual data volume is very large, the DEMO data volume is small, and the performance problem cannot be exposed.
In addition, DEMO does not have a dimension table association, if you add a dimension table association, the two main tables cannot establish a many-to-many bidirectional relationship (in a circle). Try to use the virtual relationship linkage of measures, either slow or not effective (treatas, intersect, crossfilter).
Please see if there is an optimized solution, thank you.

 

pbix file is here:

call for help.pbix

[lbendlin]

What is the business question you are trying to answer?  Number of machine codes shared across categories?

but the performance is slow

Please quantify.

[Jackson123]

hi, Ibendlin,

you are right.

the desired effect screenshot is the reuslt wanted.

when viewer clicks the upper bar chart "On" category, then the lower bar chart will highlights other categories crossfiltered parts, e.g. De highlights 588, Pr highlights 647, these are crossfiltered part (machined codes) with On category .