From the course: Machine Learning with Logistic Regression in Excel, R, and Power BI

Introducing AI and logistic regression

From the course: Machine Learning with Logistic Regression in Excel, R, and Power BI

Start my 1-month free trial Buy for my team

Introducing AI and logistic regression

“

- [Instructor] Artificial intelligence or AI, as it's often known, uses computer algorithms to replicate tasks normally performed by humans. Examples of these tasks include activities like visual perception, speech recognition, and decision-making. Statistics measures the relationship between input variables and known outcome labels for existing data. In AI algorithms, we want to predict the outcome labels for new data, using the statistics we calculated from the existing data. Machine learning is a type of AI algorithm that uses feedback loops to continuously update the model. Types of machine learning algorithms include classification, regression, clustering, and anomaly detection. Regression modeling lets us solve parameters representing the relationship between the inputs or independent variables and the outcome label, or dependent variable on existing data. We can then use these estimated parameters to predict or forecast outcome labels on new data. There are several different types of regression models, but two of the most common are linear and logistic regression. Linear regression shows the relationship between the independent input variable and the dependent variable for the outcome with a straight line, while logistic regression displays this relationship with an S shaped sigmoid curve. The outcome labels or dependent variable in linear regression are continuous numeric values. The outcome labels and logistic regression are also numeric, but they're discrete instead of continuous, which means we'll see whole member predictions instead of decimals, for example. There are three main types of logistic regression, binomial, multinomial and ordinal. Binomial logistic regression is a classic linear log model with two outcome labels, zero and one, for example. Multinomial logistic regression also has discrete outcome labels, but instead of two outcomes, it has three or more potential outcomes, but the order of them doesn't matter. Ordinal logistic regression is another type of logistic regression model with three or more possible outcomes, but order matters in these models. Examples of real world problems that we can apply logistic regression frameworks to include credit scoring, medical testing, and gaming.

Contents