Categories

# Prediction

The task of Prediction helps answer the question: “what will happen in the future?”. In a way, this is a task of oracles and fortune tellers. However, it’s better to rely on data than magic to predict future outcomes. With that in mind, Prediction becomes a building-block of the whole Machine Learning process. Predict what will happen is essential for the goal of going From Data to Decisions. It provides some guidance about the future and helps to make the right decisions in advance.

The general predictive model gets the features as input, then gives back the target as output. Going from inputs to prediction targets is Supervised Learning, since the labels are known. Generally, when the target is numeric, it’s a Regression problem. When the target is categorical, it’s a Classification problem. For both problems, there are not only traditional models, like Linear Regression and Logistic Regression, but also advanced Deep Learning techniques.

## Regression

Linear Regression is one of the most popular models in Statistics and Machine Learning. It’s a Supervised Learning method that gets attributes as input, then the model predicts a numeric target as output. The goal of linear regression is to find a linear function that minimizes the distance between observed data and data predicted by the model. To accomplish the task, the standard approach is to use the traditional Least Squares method.

One very important thing to look for is the evaluation of the model, that is how well it performs. In Linear Regression, a way to do that is to analyze the R-squared coefficient. To simplify, R-squared tells the proportion of the variance in the target that is predictable from the features. Linear regression

## Classification

Classification is a supervised learning method, just like Regression. Both Classification and Regression aim to predict a target based on some attributes. However, unlike Regression, the target of Classification models is a class, not a number. Classification is crucial for many areas, such as Credit, since it’s important to classify the customers that apply for a loan.

A simple, yet very powerful Classification model is Logistic Regression. The model receives attributes as input and predicts a class as output, simliar to Linear Regression. Although, the function that the model applies is not linear, it’s a Logistic function. This function allows the model to predict the class target. Logistic regression

### 4tune

At 4tune, we are passionate about solving clients’ problems to deliver profit. If you have a business idea or need help with anything related to data science or artificial intelligence, we are glad to help. Please contact us at victor@4tune.ai