By Garth Liebenberg – ERP Technical Services Consultant.
One of the buzzwords floating around these days is Machine Learning, but what is it and how does it apply to your enterprise?
Machine Learning is a type of artificial intelligence that enables computers to learn or teach themselves. This is different from the traditional rule based software where intelligence is explicitly programmed.
Let’s look at an example comparing the differences: A customer review site wants to automate the classification of the reviews they receive.
In the traditional rule based software you would need feedback data that you can write rules for, say a star rating. Then you could write a rule that says if a 5 star rating was selected, then it is a good review, if 1 star then a bad review and so on. In the machine learning approach the actual text of the feedback can be analysed and to determine if it was a good or bad review. Think of it as teaching a child. You teach the child, then an exam is written and the results will show if the child has learnt how to solve the problem. For Machine Learning the software is the child, it is taught with training data, after which it is tested to determine if it has learnt correctly. This process is repeated until a model is determined that can solve the problem accurately.
Some existing examples available today include:
- Image recognition – A factory worker can take an image of a part that needs to be replaced, the part is determined from the image and order sent through.
- Spam filtering – Identify emails as spam.
- Resume matching – Automate identification of the best candidates for a job.
- Search engines – How the top search results are determined.
Now that we know what it is, when should we use it?
Application in the Enterprise:
As with any problem-solving tool, Machine Learning is best suited to specific use cases. Machine Learning is best suited for problems where the scale is large and the rules are complex.
As a guide, before you decide to use Machine learning it is best to ask yourself the following questions:
- Do you need machine learning?
- Do you need to automate the task?
- High volume tasks with complex rules and unstructured data are good candidates
- Can you formulate your problem clearly?
- What do you want to predict given which input?
- Do you have sufficient examples?
- The more data, the better
- Does your problem have a regular pattern?
- Machine learning learns regularities and patterns
- Can you find meaningful representations of your data?
- Machine learning algorithms ultimately operate on numbers
- Example would be word frequency in a customer review
- How do you define success?
- Given an input, the result was predicted 90% successfully to solve a specific problem
If the answer is yes to all 6 questions then Machine Learning is the right solution for your business problem.
For more information check out the following useful links: