International Data Corporation (IDC) forecasts that expenditure on Artificial Intelligence and Machine Learning will grow from $12B in 2017 to $57.6B by 2021, while Deloitte Global predicts that the number of machine learning pilots and implementations will double in 2018 as compared to 2017, and double again by 2020.
Machine Learning potential seems to be endless. We interviewed Ana de Prado, Machine Learning Program Leader at Terminus7, to talk about the impact of this technology on business, its functions and some real-world examples of machine learning applications developed by Terminus7.
What is Machine Learning and how does it work?
Broadly speaking, Machine Learning is the subfield of Computer Science and a branch of Artificial Intelligence that is aimed at developing techniques that allow computers to learn.
More specifically, it is about creating programs that can generalize forms of behavior from information provided in the form of examples.
How advanced is this technology today?
Although much progress has been made over the past few years in Artificial Intelligence, there is still a long way to go before we achieve a mature technology.
Actually, little progress has been made in relation to the algorithms used, which are based on mathematical models created years ago. The main difference lies in the progress made with hardware and shorter computation times, a useful progress that will allow us to process the vast amounts of data currently being stored.
What types of business problems can Machine Learning handle?
Machine learning has a wide range of applications, including search engines, medical diagnostics, fraud detection in the use of credit cards, stock market analysis, classification of DNA sequences, recognition of speech and written language, games and robotics, etc.
Broadly speaking, this technology can use as many applications as you can imagine because it can adapt to the number of situations associated with the volume of data you are using.
We also talked about some of these applications in “What is machine learning and what are its main applications?”, where we mentioned some of the activities of our daily life that are driven by machine learning.
What are its main advantages?
Experts state that machines cannot act like the human brain. However, when we speak about gigabytes, terabytes or even petabytes of information to be processed for decision-making purposes in short periods of time, human beings cannot compete against machines.
Machine learning allows an efficient use of resources and time cycle reductions, which is an improvement in terms of quality.
What do we need to develop a good ML solution?
The first thing we need is the data to be trained. This data should meet a series of requirements:
- The information supplied must be sufficient
- Data must be clean and clearly labeled
- Data must be properly balanced
When the data meets these requirements, we have to think about what model would be the most appropriate to solve the problem. Next, we have to choose the libraries and frameworks to use and check the customer’s specifications.
Is feeding it with data enough?
It depends… Don’t forget that there is more than one type of ML algorithm. We could say that they are divided mainly into three categories:
- Supervised learning
- Unsupervised learning
- Reinforcement learning
Each of these categories requires a different prior treatment of the data and some (such as Reinforcement Learning) need follow-up during the training phase.
It is also necessary a validation work of the trained models, a process in which it is verified that it works correctly and the metrics of success in the predictions are obtained.
Clean data or Big Data, which is the best option?
The amount of data that is currently being generated in companies is increasing exponentially, but extracting valuable information from such data is an effort that cannot be underestimated.
In other words, data is relatively easy to gather but “Big Data is not only about the data”, its real value is associated with its analysis and it is precisely for this reason that the use of this technology in business is booming.
What kind of projects are you currently at Terminus7 working on?
We work worldwide on many different projects: Marketing Campaign Optimization, Train Tracking, Infrastructure Monitoring, Electricity Generation Prediction, Planet Detection, Call Center Classification, Churn Rate Prediction… You can explore these case studies on the Terminus7 website.
In addition, we conduct research projects with reinforcement training algorithms and research projects applied to robotics.
Do you want to know more about us?