Predictive analytics is defined as the process of using data analysis to make predictions based on that information.
Health care systems will use predictive analysis to also support decisions made in different care centers. Healthcare analysts use predictive analytics to "prevent patient illness, avoid penalties, and reduce costs"
Predictive models apply known results in order to train the model to predict values, with different or completely new data, in a repetitive process. The modeling provides the results in the form of predictions represented by the degree of probability of the target variable based on the significance estimated from a set of input variables.
Structured data is data that can be stored, consulted, analyzed and manipulated by machines, usually in data table mode. Unstructured or unstructured data is the opposite. Structured data is the classic data of the patients (name, age, sex ...) and unstructured data are paper prescriptions, medical records, handwritten notes from doctors, X-rays, CT, etc.
Healthcare
In healthcare, predictive analysis is used to determine which patients are at risk of developing certain disorders such as diabetes, asthma or cardiovascular diseases among other recurrent diseases.
The study of infectious diseases is often based on mathematical epidemiological models that attempt to emulate the dynamics of the disease and estimate the parameters related to it, such as the reproducibility rate, the mortality rate, and so on.
The basic form of this type of simulation is SIR models based on the assumption that the population can be classified into three independent compartmentalized groups (susceptible, infected and recovered person). The number and type of compartmentalized groups can be modified to better reflect disease-specific dynamics, as in SEIR (susceptible, exposed, infected, and recovered) models. The models study how individuals can progress from one compartmentalized group to the next.
In the adjustment, we determine which mathematical functions and parameter values are compatible with the data that we observe on the daily evolution of the epidemic in the past, also incorporating biomedical knowledge about the natural history of the disease, in a Bayesian approach. Once the model that best fits the data from the past has been chosen, the one that best predicts it is used to simulate or predict the evolution of cases in the future, under different intervention scenarios.