Applying Machine Learning with Decisions
Barely a day goes by without some incredible story regarding the advancements being made in machine learning. Aside from beating humans in Jeopardy, Chess, and Go, machines have also advanced incredibly fast in areas like image recognition, translation, transcription, and speech to text. Self driving cars are in the news daily as are stories about AI’s writing poems, music, and sorting cucumbers.
If machine learning can do all these things, how can we utilize machine learning within our companies to improve performance?
Actually, the tools are readily available today and can quickly be applied to some common business problems. I will be conducting a webinar this coming Thursday, March 8th to talk about applying machine learning with Decisions.
Decisions was designed specifically to quickly and easily integrate into other applications and services, all using our no-code visual design tools. The services available today are incredible from companies like Amazon, Google, and Microsoft. These services can be quickly connected to applications within Decisions and used to improve business processes.
How can a machine learning model help? Fundamentally, creating a machine learning model is creating a prediction model. They are quite easy to build using the above tools (& others) and I will walk through a basic credit risk example during the webinar.
In terms of how these can be applied in a business context, the applications are endless and even a basic model can be quite helpful. Here are few areas where machine learning can help to reduce non value-added work:
- Categorization Models (what is the category of …..)
- IT Help Ticket
- Risk profile of applicant
- Customer segment
- Scanned Document
- Numeric Prediction Models
- Probability of default
- Probability of over time/budget
- Factory/Process (Output / Quality )
- Supply Chain Items shipped on-time
The key to implementing these models lies in what you do with the model output. This is where Decisions workflow and rule engine really helps. At the end of the day utilizing machine learning and applying it to your process is always a combination of bottom up (machine learning) and top down (rules & workflow processes) working together.