Let’s talk about Advanced Analytics: A brief look at Artificial Intelligence

In recent years, the use of certain terms to refer to new technology trends has become normal. We can find plenty of news on artificial intelligence (AI), machine learning, Big Data and advanced analytics, for example. But what does each of these terms mean? And, what do they mean in a business sense?

Over time, I’ve frequently found that there is some confusion about what these technologies cover and what they do. With this post, I will try to explain these concepts, all of them linked.

Artificial Intelligence

Artificial Intelligence is an emerging term that has created a growing dialogue among businesses leaders and prosperous niche, appearing startups and solutions based on AI. Some well-known examples of products based on AI include recommendation systems, chatbots and self-driving cars. We can find multiple instances of solutions based on AI present in our day-to-day transforming the ways businesses operate.

But what do we understand by AI? Artificial Intelligence is a broad science concept that describes the ability of machines to make smart things based on logic processes. How is it created? To develop AI, experts use software that allows AI to make conclusions at the same it processes smart data.

The concept of AI dates back the imagination of human beings from ancient Greece, and the first computers were logic machines with software intelligent enough to solve basic calculations. As we have increased our understanding about the human mind, the definition of AI has evolved. Nowadays, we can say the concept of AI is the ability of machines to imitate the decision-making process of human beings and execute smart tasks as human being would.

Human intelligence is made up of several capacities: learning, reasoning, problem solving, perception and the ability to understand language. All of these involve analytic skills. AI solutions also use analytical skills. In the graph below, we can see the different levels and types of analytics, from the most basic to the most sophisticated.

The basis of all analytic activity focuses on understanding what is happening using historic information. This is covered by descriptive analytics, which helps to understand what happened, and diagnostic analytics, which try to understand why it happened. From business perspective, these methods have traditionally been covered by Datamining and Business Intelligence.

The next level is covered by three disciplines known as advanced analytics. They try to use and add knowledge to make better decisions in the future and they also provide additional insight. This is what we understand nowadays as AI.

Each of these disciplines use statistics and math techniques to achieve the proposed goals. They help us understand what and why something happens, what could happen, and what someone could do to make something specific happen. Additionally, these techniques also focus on how to make decisions like humans would. The AI technologies can automatize decision-making by guaranteeing the quality of these decisions and their contributions to business.

All of them need large amounts of data for make this possible. Here is when other terms come into play. Perhaps you’ve heard of one of them: Big Data.

Therefore, AI makes sense if we have data that can be mined through analytics technologies to achieve our goals. These goals can help us: getting better knowledge about our customer, improving the process efficiency and increasing the profitability of services. This reason place value on this concept acquired in business organizations. This is the reason why business organization place value on AI.

Big Data

This concept focuses the activities related to systems for treating and handling large amounts of data. Big Data involve using enormous amounts of data that we need to fulfill our goals, as noted before. Big Data supplies the infrastructure and capability to capture, store, treat and transfer this large amount of data.

Big Data allows experts to handle unstructured data from any source (traditional data base, social networks, IoT devices, etc.), any format (image, audio, video or text) in real time.

These are the five defining concepts of Big Data:

Well, we are going to focus on these disciplines around advanced analytics, and zoom in on each of their techniques.

Predictives Analytics vs. Machine Learning

Machine learning has been used extensively by data scientists, business managers and executives for some time. There is some confusion surrounding the concepts of machine learning and predictive analytics. Which do these terms mean? And, what is the difference?

Both disciplines have the same objectives: forecasting. The difference lies in the amount of data involved and the human participation during predictive models building.

Predictive analytics uses statistical techniques for evaluating behaviors and determining whether a result is viable. This means, trying to address what can happen in predicted future situations. Its nature is probabilistic; because it tells us what is the probability that something will happen. We can do this by trying to find relationships and patterns between variables by using present and historic information. This way, we can extract conclusions and prediction through data. The most well-known applications of Big Data are the credit ratings used by financial companies to evaluate probability of future defaults. There are many cases, such as a customer’s calculated propensity of purchase to acquire a product, churn scoring, etc. The greater the amount of data, the more accurate the information is.

There are different modeling statistic techniques (as shown in the following graph). Choosing one depends on the necessity of having a descriptive analysis of result for understanding why take a result, the types of data and their structure.

Machine learning is a discipline that tries to build complex models and algorithms, in order to create prediction. This make it different with other cases because it does so without programming explicit instructions, and data learning, allowing models to evolve and adapt while they add new data. That is, we only work with data available and the desired result.

The result of machine learning is a prediction that can guide decisions in real time without depending on human intervention.

The main approach includes the used of neural networks, generic algorithms, rules induction and analytics learning. But, in contrast to predictive analytics (which the different models use independently), machine learning is hybrid, mixing different types of models.

The combination of analytics models can guarantee efficient results, repeatable and reliable. This is particularly appreciated in business solutions.

Many times we can see this discipline associated with cognitive computation as we will discuss later. This is because one of the applications of machine learning is the automatization of knowledge acquisition in systems that intend to emulate the decision-making process by human beings.

There are cases with great impact. For instance, Amazon uses it for customized recommendations of products based on behavior surfing and the customer shopping experience. Google uses it to improve search results and Facebook uses it in image recognition, labeling and connecting known people in pictures. In addition, we can check other business applications as shown in the following graph:

I can find a simile that I have seen well suited in SHARP SIGHT LABS. He tells us that, ML and predictive statistics modeling are identical twins. They are nearly identical. If you compare them, they have the same DNA like twins. However, they are different, relating with different people and behaving distinctly.