# What Is Predictive Analytics?

Predictive analytics uses statistics and modeling techniques to predict future outcomes using existing data. You can use predictive analytics in various applications, including weather forecasting, video game development, customer service and investment portfolio development.

Companies also use predictive analytics to identify risks and opportunities in data to drive strategic decisions. The data used for predictive analytics can come from a variety of sources, including log files, databases or Internet of Things (IoT) sensors. This type of data is valuable, however it often suffers from low quality and inconsistent structure, which can be challenging to work with. Predictive analytics tries to solve these problems by using machine learning algorithms to build models that are then used to make predictions.

An example of predictive analytics is an automated system that predicts the risk of injury in sports players based on their physical characteristics and performance history. A system like this could help coaches decide when to take a player out of a game or prevent injuries by adjusting training regimens. It could also help athletic trainers determine whether treatment is necessary before heading to the emergency room.

You could also use predictive analytics in healthcare to predict how diseases might spread based on patient symptoms and other factors. By predicting the likelihood of an outbreak before it occurs, the health department can prevent it from spreading further and take control of a situation early on.

### How Predictive Analytics Works

Predictive analytics works using a concept called modeling. A statistical model is a set of assumptions and relationships used to represent the real world. In predictive analytics, a statistical model describes the relationship between a dependent variable (the thing you want to predict) and one or more independent variables (items that might affect the dependent variable). For example, you might want to predict whether a customer will respond favorably to an advertisement. In this example, you might use customer age, income level and gender as independent variables in your predictive model to predict how likely a customer will respond favorably to your ad.

### Modeling Approaches in Predictive Analytics

There are two high-level statistical modeling approaches:

#### Classification Model

A classification model analyzes the relationship between variables (or features) and a target variable (or outcome). For example, a doctor might want to know whether or not a patient has cancer. The doctor can use a classification model to determine if the patient’s age, gender, height, weight or other characteristics may be related to the presence of cancer.

#### Regression Model

A regression model analyzes variables’ relationships and interactions. For example, a doctor might establish that a patient has depression and insomnia concurrently. The doctor could then use a regression model to determine if there is a probability that these two symptoms are related. A regression model is used in this instance to tell us how much one system affects the other.

There are two main regression models:

• Linear Regression: Linear regression models use a straight line to model the relationship between dependent and independent variables. This approach is useful when there is a specific period in which the relationship between variables can be observed—for example, predicting sales based on weather data. The dependent variable is referred to as the response variable. Linear regression predicts a dependent variable typically associated with the independent variables. In some instances, linear regression models can also be used to predict values of an independent variable that are known to change over time.
• Non-linear Regression: You would use non-linear regression models when there is no clear linear pattern between two variables through a single line (for example, the sales of products). Instead, these types of models use multiple lines to predict how one value of one variable will influence another value of another variable. Using multicollinearity analysis, these models can also be used to simultaneously predict values for multiple dependent variables or analyze multiple independent variables.

### Techniques for Building Statistical Models

Data scientists use several predictive analytics techniques to construct classification and regression models. The most common are:

• Decision Trees: Decision trees are hierarchical tree structures, which consist of a root node, branches, internal nodes and leaf nodes. This approach classifies objects into different categories based on the value of the independent variable at that node. Each new node in the tree corresponds to an additional level of dependent variables. Decision trees are helpful for regression models and for building classification models.
• Neural Networks: Neural networks consist of layers of interconnected artificial neurons. The first layer of neurons receives input data and converts it into a form that the rest of the network can understand. The final layer of neurons generates the output as a prediction. The nodes in each layer are connected with a weight value, which can be adjusted to improve accuracy as more data is collected. The network’s output is compared with the desired output value, and the difference between the actual and expected output helps determine which classification is likely correct. Predictive neural networks are often used in real-time applications, such as stock market analysis, risk assessment or fraud detection.

### Predictive Analytics: The Key to Data-Driven Decisions

The more data a company collects, the more value it can provide. Predictive analytics is key to uncovering hidden patterns to unlock the value in this data. Predictive analytics helps companies make informed decisions, communicate better with customers, and improve operational efficiency.

### Your Data Science Journey Starts Here

Predictive analytics is a valuable skill for data scientists, and you can learn more about the different models used in predictive analytics and how they can be applied to solve real-world problems with The Data Incubator.

The Data Incubator is a data science bootcamp and placement company that offers training in the latest data science specialties.  Our programs prepare students for new career paths, advanced education and skill refinement.  We partner with leading organizations to place our highly trained graduates. Our Data Science Bootcamp and Data Science Engineering Bootcamp give you the skills you need to excel in the field.

If you want a more immersive experience, check out our Data Science Fellowship Program. This 8-week bootcamp provides a deep dive into data science concepts through live coding and real-world data sets.