Why Python and Power BI?
Nowadays, reporting solutions (Power BI, Tableau) are mentioned as BI - business intelligence -, however they just summarize, transform as visualize the data. Which is really helpful to understand it by applying measures, charts, KPIs etc. On the other hand, this setup is only for analyzing and understanding, but unfortunately without intelligence in my view.
Where does intelligence comes into the picture? For example, when we can see correlation between data and we can make predictions and/or decisions based on that. Naturally, BI tools not really capable to do this by their own setup and embedded features.
Several years ago, only server based applications were available applying mathematical models and statistic calculations. Developers put together something on the database server, and users could see some tables with predictions.
With Python integration of Power BI, it become easy to check correlation of your data. Imagine, a scatter plot is applied – as a standard feature of Power BI. Let’s see if there is a correlation between the points - the various data. For example, python can do linear, polynomial, an multiple regression analysis easily by built in modules. Also training and testing the models is possible. With scoring, models can display if there is any correlation and how good the model is.
Scatter:
Regression:
Or, check if there is a more complex relationship between your data elements. For example, there is a targeted mail list for bicycle buyers. You would like to see if there is a relationship between age, distance from home, yearly income, children number, gender (factors) and if someone a buyer or not (result). With Power BI and Python, you can easily add to your report a decision tree calculation algorytm to see if there is any correlation. Also, with slicers you can parametrize the prediction command, so you can use your dashboard for scenario analysis.
Intelligence can be added by 3 easy steps to the data:
- selecting the measures in Power BI correlation should be investigated
- adding Python visualization and typing not more than 30 lines of code
- plotting charts and based on model scores decision making and predicting
This tool combination gives a powerful asset to help data analytics and decision making, and it is fully available in a desktop software. Hopefully, the more data and scenarios we need to work with, the wider range this application will be used.