Original article was published on Artificial Intelligence on Medium
A friendly Machine Learning concept !
Before to talk about machine learning concepts, let’s talk a bit about how much the electronic devices have evolved, and some cool features we can found.
For example, nowadays your cellphone it’s able to recognize your voice; further more, you can give orders to it, and it will do it. If you really think on it, your cellphone it’s a very intelligent device, and that is the mainly reason why the “smart-phone” concept was born.
But let’s think deeply about smart-phone concept, as you have seen, there are apps (applications) for image editing which can recognize your face and let you do some adjustments without modify all the scenario or contour. There are also apps, that let you use the camera in real time to superimpose a hat in your head, it doesn’t matter your body position, they’re are able to recognize your body and dress it with clothing articles images; yes, the current technology is simply amazing !.
Analyzing a robot.
Let’s think into a robot, they doesn’t have feelings like us, instead they have lot of “sensor” devices integrated to read in digital language what’s happening on the environment, but that’s not all. We are the ones that can program the robot to tell it how it must process the lecture of all those received signals. That’s the way in which we tell to a robot how it must “think” about everything, and also how it’s reaction must be.
Maybe you want to take a look of this video about Sophia Robot, it has been built with all resources of Artificial Intelligence (also it works with Machine Learning Algorithms), you will be surprised with this video conference !.
Robot’s visual sensors with the correct programming can do lots of things:
- Identify your emotional state (with face recognition and body positions).
- Identify when you’re trying to remember something (neural linguistic human concepts about head and eyes position when you’re thinking).
- Insecurity detection or lack of knowledge about something.
with audition sensors, we can also teach to an intelligent device how to recognize voice patterns, or convert phonemes into text that will be used later on by the device itself to “understand” or “do” things in response of it.
What about Machine Learning ? What’s that…?
Yeah you’re right, it’s not all just about sensors and how to read the environment trough them, today’s advance electronic devices like the smart phones, laptops, use a technology called Machine Learning, which are basically algorithms that are continuously saving statistical information about you, let’s said, how many time’s you check the weather, how many time’s you ask to your voice assistant about a same thing, or how many times you interact with your phone apps to suggest you in a couple of weeks which of them you’re not using and maybe you could erase to optimize your phone’s memory !.
Machine learning is the ability that has an electronic device to learn things about you, they are collecting data all the time and evaluating it according to big data statistical collection. This way they can have an idea about how you’re thinking, how you’re feeling and what really likes you, etc. The machine learning programs or algorithms are all the time studying behavior patterns, they can predict things according to the collected data, and they can give you very accurate analysis.
Yes, Machine Learning are processing data all the time, and as fast as learn about you, they can suggest you important things, or remember to you things like meetings or appointments, or recommend you to check some news taking into account your preferences, or recommend some new songs, etc.
Some examples of Machine learning uses:
- Machine learning is behind recommendation of movies in big digital platforms like Netflix or Spotify (of course they’re recording and analyzing the users statistical interactions, and also your personal interactions).
- In autonomous cars, it’s the ability they have to recognize the streets, and paths to move into, the data has being collected, processed and stored all the time.
- There are smart-phone applications, that can identify an object or a place and give you accurate information about it, name of the object, or web pages in which you can buy it, where is located, etc. (Google lens — see this link).
- Chat bots are example of machine learning technology implementation, they use this technology to understand the users words and to predict the most accurate answers, a bot not only take the words, it tries to imitate the thinking process of a human, so words can have a very clear sense.
- Machine Learning helps in medical purposes due to all it’s data records about symptoms and diseases, letting to users (doctors) obtain very accuracy diagnostics. IBM Watson’s artificial intelligence can help to solve complex medical cases ! < read it here in this article >.
- Machine Learning It also has being recognized in worldwide news, due to it’s ability to win games such as go, or Rubik cube challenges.
- It’s used to predict when technological equipment it’s going to fail, and also can suggest you when you should purchase of your critical spare parts for your local stock.
- It also can be used to know which employees will be more cost/efficient on next year ! (Yes, human resources departments are making bets about the use of Machine Learning).
- It has being used in social networks, or trough websites, to select customers with high purchase potential.
- In cities, it has being used to predict urban traffic, and to suggest how to reduce considerably your travel time (Example: Google Maps).
Machine Learning is a branch of A.I. (Artificial Intelligence) that let the machines to “learn by itself without have being programmed to do so”. As we have said before, this technology can identify patterns in all the data to do accurate predictions. This kind of prediction is also used to improve motor-search websites, like google, it’s really helpful for all the users !.
When we refer to “MACHINE LEARNING” we’re really talking about an algorithm or program, that checks all the data and basically make future predictions, also Machine Learning concept implies that all these systems are improving all the time by his own way, without human intervention.
Machine Learning applied to the companies.
Searching on the web, we found a very interesting article of cleverdata.io, in which, they explain a simulated example about how the Machine learning concept can be applied to companies to predict bad business scenarios, giving to that companies the option to change the course and goals, to get succeed.
You can find the whole article here, but let’s do a short summary:
1. The scenario:
A telephonic company that wants to know which customers are canceling their services and why, so they can implement some strategies to avoid that customer’s loose.
2. The question:
How this company can do that ?. — They have lot of data from the customers, call registers, type of services contracted, diary consumption, number of calls done to customer support, last changes to contracted services. etc. (very interesting data).
3. The solution:
If there exist all that user data, then MACHINE LEARNING can be absolutely implemented to that company, to make the problem turns from REACTIVE to PROACTIVE !.
4. The implementation:
That big amount of data can’t be analyzed by a human, and even it will be hard if want to do predictions concerned to it. BUT ALGORITHMS instead can easily detect behavior patterns in big amounts of data, taking into account all the variables in the data. The Machine learning can predict which of those variable are the responsible of the customer’s loose.
5. Tools to implement Machine Learning:
These are the most popular Machine Learning software tools to create our algorithms, according to this Article posted on June 30th-2020:
Scikit Learn, PyTorch, TensorFlow, Weka, KNIME, Colab, Apache Mahout, Accors.Net, Shogun, Keras.io, Rapid Miner.
Please check the article, to see the comparison chart with the summary of features for every algorithm.
6. The answer of Machine Learning to the company:
This is the detected pattern for the users that cancel their contract:
* The user has more than 3 calls to customer assistant support.
* The user calls less than 171 min per day.
* The user calls during night lasted less than 189 mins.
According to the Machine Learning algorithm, the active customers with a similar pattern has a 91% of probability to cancel the contracted services with this telephonic company.
Why it’s important Machine Learning nowadays ?
If we think about it, current companies generate lot of data (Exponential growth), and they are still incrementing data. those companies can extract valuable information from the data, and if they do so, then will have a big competitive advantage !.
Also with internet/networks revolution (Internet of things), the quality of the data it’s very high. Therefore the predictions have a very significant value to take decisions or develop new strategies for the company. This is the main reason why the data should not be ignored.
it’s just the perfect opportunity to focus the company attention, you don’t have to be an expert to approach this type of technologies, also in the market you can find easy use Machine learning tools and with reasonable prices.
Today we can have a Machine learning high qualities results, it’s normal that you think that if have more amount of data the results will be better. That’s not necessarily true. You don’t need as many data as a platform like Facebook or such a big bank company, you have to focus your attention into obtain high quality data (trust data) instead of having millions of data with a very low reliability level.
In Summary, when it is possible to implement Machine Learning ?
According to all the things mentioned before:
- It’s a fact that you must have a considerable amount of data, and a periodical collection/flow of them.
- Then you have to be sure about the possibility to extract valuable information from all these amount of data.
- Then you have to identify the quality of that obtained data (you have to trust in that data).
- Once you obtain the results from a Machine Learning algorithm, you must take into account it’s predictions, and do something with results !.