What is the Data Science, what it does and with what objectives

With the term Data Science means a discipline that deals with processing and interpreting data. Discipline Data Science it arises from the convergence of many knowledge ranging from Statistics to Mathematics, from Science to Computer Science. 

La Data Science, or Data Science, is a relatively new science, in fact it has been around for about fifty years. It arises from the need to put order in a very lively and rapidly evolving context. The growth in the volume of data, the possibility and ability to give meaning to data, have made the Data Science.

Historically speaking, data has often been treated as a kind of secondary product of any process. Anyone over the centuries has undertaken to collect data, has done so mainly for their own convenience, often without imagining that today an economic value could be attributed to a collection of data. If we think, for example, of a farm that over the years may have collected information on crops, events, sowing, etc., perhaps it could have done so to archive its corporate history. If all farms had done that method, then fertilizer companies today could benefit from it for research purposes, or for marketing purposes.

Data Scientist

The one who deals with Data Science, he's called data scientist: currently one of the most sought after professionals in the world of work.

The task of the data scientist is to analyze data in order to identify models within them, that is, what I express the data available through the trend. The identification of these models is functional to the client's purposes: company, public body etc ...

Monetization

In recent years, a data marketing model has increasingly established itself where someone is interested in selling data and someone else in buying it.

Companies specialized in the production of data were born, and companies specialized in buying and selling after appropriate cleaning and reprocessing operations. If we then think about the privacy regulations, we realize the complexity of the subject. Today there are strict laws that call for a conscious and respectful use of information.

Planning

A project by Data Science usually consists of the following steps:

1. Description of the problem: a project is born with the aim of achieving a result, or solving a problem. This description defines the characteristics of the results to be obtained, the timing within which to achieve them, and the resources available;
2. Data collection: to tackle the problem, datasets will have to be obtained, either already present in the company or from external sources. This could be online surveys, data collected from social sources, etc…. This dataset will then be cleaned up, ordered and structured in such a way that it becomes the source of work of the data scientist;
3. Exploration of the starting data: at this point the data must be viewed and explored, filtered, summarized in tables and displayed in graphs. All this will make it possible to identify its characteristics, boundaries and any gaps; 
4. Data analysis: this is the phase in which the models are identified, using the most suitable tools and techniques according to the objectives to be achieved;
5. Application of the Models: in this final phase, the application of the identified models will allow to obtain the desired results. Here the data scientist it merely gives guidelines for the use of the data set, and the extrapolation of the correct information. These guidelines must be expressed in the company's business language, which will serve to solve the problems for which the project was developed.

In every single step the data scientist interacts with specific company departments, and therefore we can say that the data scientist is perfectly integrated into the corporate reality. 

With technological advancement, the data scientist he has often found himself facing problems of Big Data and Artificial Intelligence.

Big Data

When we talk about Big Data we refer to data that contains a great variety, arriving in increasing volumes and with greater speed. This concept is also known as the rule of the three Vs, which consists in the choice of three terms that characterize the Big Data phenomenon in its essential features:

1. Volume: because the amount of data to be processed is extremely high. This factor is certainly the most indicative of the three, in fact the word Data is preceded by the adjective Big. Big Data archives are measured in the order of Terabytes or Petabytes;
2. Speed: This term refers to the speed with which data is accumulated. The contexts in which the flow of data is continuous, non-stop, streaming flows onto the server are increasingly common. This forces the implementation of systems capable of accumulating without data loss, without blocks and minimizing latency times. Let's think, for example, of financial transactions, online trading, IoT etc ...
3. Variety: unlike highly structured archives, the data that is part of a BigData system is very varied. As an example we can think of the data that is collected in any social network: photos, text, attachments, videos.

In reality, other peculiarities have also been added over time, such as the truthfulness of the data to identify the reliability and reliability of the data.

Large volume of data arriving at great speed, and characterized by great variety, necessarily lead to data organization problems.

How to organize them?

Welcoming them and then processing them? Structuring them and then processing them?

Several paradigms of organization of data systems were born, which have established themselves over time:

• Data warehouse: That is, structured data archives. Widely used, with the particularity that data must be organized at the very moment in which it is received;
• data lake- characterized by the fact that all incoming data is set aside in the container without any structured storage policy. The opposite paradigm to the Datawarehouse, because in the Data Lake the data will be structured only when it is necessary to read them in order to interpret them. This approach simplifies and speeds up the acquisition phase, to the detriment of the subsequent phases;
• Silos: also called compartmentalized. Each department or company department accumulates its data without sharing it with others.

Currently these are the most widely used paradigms, and in many cases the solution of integration prevails, ie different projects could use different accumulation methodologies and then integrate themselves at a later time. There could be situations in which different data are collected with different paradigms, or different collections could constitute contiguous phases of the same life cycle.

Machine Learning

Despite their great usefulness, we know very well that processing machines or computers are stupid. That is, a computer can do nothing if it is not the human being to analyze a problem, formulate an algorithm and encode it in a program.

This has always been the case, until we started talking about Artificial intelligence. In fact, artificial intelligence consists in inducing a kind of spontaneous reasoning in the machine, which can lead it to solve problems independently, that is, without direct human guidance.

It took several years before the expression "induce a kind of spontaneous reasoning in the machine“, That is, it took several years before we passed from a condition of total“ forced ”instruction of the machine, to a condition of self-learning. In other words, the machine has been able to self-learn, to learn. We have therefore arrived at Machine Learning.

Machine Learning is a branch of Artificial Intelligence in which the programmer drives the machine in a training phase based on the study of historical data. At the end of this training phase, a model is produced that can be applied in solving problems, explained with new data.

I respect the classic approach, where the data scientist used to work for definish solution algorithms, the machine will discover what makes up the model. The Data Scientist must take care of organizing increasingly effective training phases, with richer and more significant data, and of verifying the validity of the models produced by subjecting them to tests.

Thanks to Machine Learning, the systems we use in mobile devices, internet, home automation are (or seem) more and more intelligent. A system, as it works, may also be able to collect data on it and on the users who use it, then use them in the training phase and then further improve the forecasts.

Ercole Palmeri: Innovation addicted