How to start with machine learning in your business

The whole world is talking about artificial intelligence, algorithms to assist drivers, doctors, programmers, artists and many other industries. The largest companies introduce solutions based on a neural network to optimize internal processes, increase sales or support colleagues. The question is whether this also translates into smaller companies and, therefore, whether it is a forceful solution or actually supports the company. In this article, I will try to figure out the topic and, based on my own experience and conversations with people from the industry, determine what applications artificial intelligence has in small, medium and large enterprises.

Below, I present a series of points describing how to implement machine learning in your company, while highlighting some pros and cons of a specific solution.

1. Solve a small problem

If you do not have experience in implementing ML-based solutions, first investigate whether you have a specific problem that can be solved with ML. First, ask yourself what the fundamental purpose of AI might be. I'm in a hurry with the answer, well, the answer is prediction. Take an e-commerce store as an example, we often wonder how to attract new customers. For this purpose, we introduce proven solutions, such as sponsored campaigns on Google ADs, Facebook, Instagram, etc. This is obviously a sensible solution, but it should not be a determinant of an ideal operation. Since you run a store, you probably have a customer base and their purchases. It allows you to make predictions based on the neural network of how your clients will change over time. In addition, it gives the opportunity to introduce product recommendations and thus increase sales conversion. Now let's move on to a more detailed description in the next point. At Pluscode, before implementing solutions based on Machine Learning, we try to analyze the problem first, estimate the time of execution and the result of a given action, on a smaller amount of data.

‍

‍

2. If this is a minor problem, can Supervised Learning be able to solve it? Supervised learning is nothing more than a specific learning of a neural network based on historical data. Supervised refers to human evaluation of the results that result from the operation of a neural network. This kind of neural network allows us to:

‍

• Predict the cost of a given product
• See the dependencies between different processes
• Predict the customer's shopping path

2. Do I need a lot of data?

‍

‍
The answer is very unclear. At the same time, Big Data is not always required and is usually very difficult to obtain. On the other hand, without a certain minimum amount of data, we will not be able to find specific dependencies in the data. At Pluscode, we analyze the available data and specific problems that can be solved with AI and define ourselves as to the possibilities.

3. Local machine learning machine or cloud-based solutions?

‍
There is also no clear answer to the question here. On the one hand, having your own equipment is clear when it comes to costs, while spending a one-time budget on specialized graphics cards is also not a cheap business. Cloud-based solutions come to the rescue. They allow you to save money and pay only for the space used and computing power. One of the examples of providers of such services is Amazon AWS, Microsoft's AzureML or Google Cloud Platfrom. At Pluscode, we specialize in solutions based on AWS services, while each project is considered individually.

In fact, the optimization of processes within companies is not always obvious. Solutions based on artificial intelligence can be a bull's eye or a strenuous attempt to use technological innovations. Everything really depends and at Pluscode we make sure that such puzzles are solved in the most economically profitable way.

‍

‍