26 DEC 2017

Predictive Analytics, analyze data to predict future trends

Predictive Analytics is a mix of data analysis related to phenomena observed in the past integrated into predictive models and machine learning with the aim of predicting future trends. Modern Predictive Analysis is more than ever focused on data, Big Data, as well as large volumes of data segmentable by different variables. Today we have a lot of Big Data.

The process of Predictive Analysis consists mainly of identifying data considered significant and suitable to describe the phenomenon to be studied and then analyzing this data looking for trends or patterns which allow us to understand which portions of this data are necessary. This phase involves the realization of one or more models that are used to describe the phenomenon. Then these models become the basis for the real Predictive Analysis.

Predictive models, descriptive models and decision models: differences and interactions

There are differences between prescriptive, descriptive and decision models. An example to explain the difference between models can be the following.

Suppose you have a used car shop and want to estimate the price of a used car. Based on the prices of used cars that we have sold until now, we can build a model, a descriptive model, which relates the characteristics of the car (brand, model, mileage, general conditions, accessories) to the price. Using this descriptive model, if someone wants us to buy their used car (to be resold), we can evaluate the price at which we would be able to sell it (a predictive model). Finally, a decision model is what helps us to make decisions, once we have understood the phenomenon: in our case, a simple decision model could be “if the price we pay for the car is at least $1000 less than the price we estimate we can resell it, we buy it; otherwise, we don’t”.

The three models, used in sequence as shown in the example, are a good example of how these models work together to estimate or forecast at the business level. But what can the applications be of Predictive Analytics for companies?

Predictive Analytics applications can be used in different fields, from marketing to prevention of insurance fraud, from estimating operating risks to estimating stock market values.

Predictive algorithms and Big Data, dependencies and possibilities

Predictive algorithms base their effectiveness on the availability of quality data, possibly in quantity. Thanks to Big Data, the value of seemingly unrelated phenomena can be assessed, thanks to the data available.

In a famous study, Google was able to foresee areas where influenza spread in advance of the World Health Organization, simply observing how frequently users, divided into their respective geographical areas, looked for terms related to the flu, like “flu symptoms”, “flu remedies” and “flu medicine”.

Having significant volumes of data and the ability to analyze them, allows you to have the necessary foundation to develop realistic and reliable predictive models, to be evaluated in relation to the reference context and the source of Big Data, as well as the collection of the same.

Predictive analysis, software and models for the analysis

Predictive Analysis uses large volumes of data, which are analyzed using ad hoc software. Alternatively, mathematical modeling languages or traditional programming languages are used, together with specialized software libraries, to produce ad hoc applications based on the specific needs of customers for data analysis.

The most famous Predictive Analytics software is “IBM Watson”, which also defeated human competitors at Jeopardy, a famous television quiz show in the United States.

Elaborated by Lucia D’Adamo, in collaboration with Luigi Laura, supervised by Marco Pirrone

 

Latest News