In Part 1 I provided you with several resources for learning R & Python, machine learning, and integrating it with Power BI. Now, I am going to lay out some amazing resources and examples of machine learning within Power BI.
Classification
Summary
Machine Learning in Power BI with PyCaret (Python)
Classification is one of the most common uses of machine learning and aims to assign a class label to something. One common example we are all familiar with is our emails spam filter classifying emails as either spam or not spam. This looks at another great example - customer churn rates or essentially determining when a customer will stop becoming a customer.
Forecasting
Summary
Forecasting is probably the one part of machine learning that most people are familiar with. Forecasting is notoriously time consuming, can be wildly inconsistent, and is rarely easy. Machine learning can automate the forecasting process, optimize forecast models, and improve the performance. In the hands of someone who knows what they are doing, a machine learning forecast model is magic.
Time Series Modeling & Forecasting
If you are unfamiliar with Radacad, I highly suggest you bookmark their website. They regularly put out some great Power BI training content and often provide some great R and Python tutorials. Their series on time series and forecasting is a great introduction to the subject.
Time Series Forecasting in Power BI (Python)
Easily one of the most comprehensive posts on the subject and a great tutorial on integrating Python into Power BI.
Sentiment Analysis
Summary
How to Enrich Power BI with Pre-Built Sentiment Analysis Models ( R )
The most commonly known use for sentiment analysis is when evaluating if a customer review is positive and negative. A few other uses for it can be when analyzing tweets, surveys, and questionnaires. If you are involved in customer service, this will definitely be in your tool belt. In the hands of a master, sentiment analysis can completely transform your business through improving customer and public relations and finding what your customers truly value.
This is a great article on how to use a sentiment analysis machine learning mode and have pre-built into a SQL Server. For those more familiar with SQL, I’d suggest starting here.
Other Examples
Summary
Modeling Infectious Diseases - The SIR Epidemic Model (Going Further with Python Visuals in Power BI)
This is one of my favorite examples out there of a real life use for machine learning and one that has real world implications. The SIR mode is an example of what is called a compartmental model and is one of the simplest models but is the backbone of many of the other compartmental models out there. The model consists of three compartments:
- S: The number of susceptible individuals.
- I: The number of infectious individuals
- R: The number of removed individuals (cured, immune, or deceased)
There are a number of other models out there for infectious diseases, but this is a great introduction and example of using one within Power BI.
Process Mining
Process Mining with Power BI and R ( R )
Process mining is one application that is one of the most underutilized and underrated machine learning applications out there. It’s not exclusive to machine learning, but in the modern world, businesses are extremely complex and it’s extremely time consuming analyzing all of that data. Process mining using machine learning aims to analyze a business processes, what causes them, if they are within ideal model, root cause analysis, determine the optimal changes to make, and improve the process from start to finish.
A great example is billing and invoicing. Process mining looks at when the order was placed, when the sales rep sent the invoice to finance, when finance processed the invoice, when product was released, when it shipped, and when it arrived. If then looks at if there are delays, backlogs, inefficiencies, or significant costs, determines the cause, finds optimal solutions, and implements the solution to either speed up your process, save money, improve profits, reduce errors, or all of the above.
I’ll be updating this post more in the future with some more examples and I am planning to post a part three with some original content that’s will be a bit more fun than some of these.
