There has been so much talk about machine learning that the awareness of people and companies of its existence has become commonplace. The knowledge that there is machine learning, and that it could even be of some use, is often the starting point for discussions. This is often also the point up to which the interested party is sufficiently informed. Soon after that you get lost in a jungle of abbreviations and terms related to machine learning.
In order to proceed from this point at our machine learning digital workshops, we usually divide the workshop into two parts. In the first part, we focus on finding out what kind of tasks are suitable for machine learning and what machine learning is in general. In the latter section, we focus on the workshop participants and their business.
In our view, the utilisation of machine learning must be business and result oriented.
In the group work that forms the latter part of our workshop, we use four auxiliary questions. The first of them focuses on business challenges. Often the first challenge is at the customer interface, where you want to utilise new technology to add new features and, perhaps, marketing value. Another challenge area is internal processes, which from a data processing and decision-making perspective often contain many processes that can be automated.
And that is what machine learning is, automation of decision-making.
From identifying challenges, we proceed to identify an appropriate machine-learning problem. The preceding, more lecture-like part of the workshop has often provided adequate material for this top-level reflection. This is the next point in using the auxiliary questions; to formulate together the identified challenges into a problem that machine learning can solve.
Our previous machine learning blog briefly touched on the data and its role in a successful machine-learning project. Although machine learning algorithms would often be the first thing you think about when considering these methods, yet in almost every project utilising machine learning and related algorithms, data collection and processing accounts for the majority of the labour cost. The data also play an important role in considering which methods to utilise. Indeed, the third of the workshop auxiliary questions focuses on the existence of appropriate data, and which data would be relevant in order to solve the identified challenge. Sometimes it is necessary to consider how to make available the data required. What kind of a workload and cost would it be to obtain it?
The last one of the auxiliary questions is the most important. As in business in general, the goal in making use of machine learning methods must be profitability and productivity. What are the benefits of the desired solution? Does it bring savings? Does it increase cash flow? How could we measure the performance of a method once it becomes part of the business? Or, to put it even more bluntly – how many euros would the solution bring?
Please read more about machine learning!