Register now to learn Fabric in free live sessions led by the best Microsoft experts. From Apr 16 to May 9, in English and Spanish.
02-21-2021 09:01 AM - last edited 02-21-2021 09:03 AM
Check my LinkedIn for more work!
https://www.linkedin.com/in/petr-slamenik-7a282990/
Marketing data analysis by Maven Analytics!
Python part:
XGBoost machine learning model (target is Response)
-Before tuning: 86% Accuracy, 51% Recall
-After tuning 91% Accuracy, 91% Recall
There is huge progress in Recall, which means the model is correct in 91% cases after tune when predicting true and false at the same time.
Composite score:
I made a composite score from Recency, Frequency, Monetary (RFM) using Python. You can see the best and the worst customers in my dashboard. More they spend, the more frequent they buy and lower recency, higher is composite score.
PowerBI part:
I wanted to make distributions, whole-to-part, cumulatives, trends heatmap in my dashboard in general.
I´ve used mainly purple, grey and black color, to get a muted, consistent look. There are few filters and a lot of DAX averages. Also i did few binnings, mainly from composite score (Very good, Good, Bad, Very bad)
Also, I´ve used relationships between tables to get continent names from country names and custom sortings.
eyJrIjoiMTcyYWVjM2YtMTY4OC00Yzc5LWE5YTEtOTg5OGI2OGI1ZjE5IiwidCI6ImQyZjljZjBlLTc0ZDEtNGNiMi1hZTk5LWRmZTYyMjkxOGQ1MCIsImMiOjl9
Great job! Could you please share the .pbix file for further evaluation? I'm interested in exploring how you approached the problem or generated the problematic statement on your own, as this is part of our exploratory analysis.