Logistic Regression for Classification.

Logistic Regression is simple to implement, easy to interpret and simple to train. This is a key component in any data scientists toolbag. Learn all about Logistic Regression, when to apply it and how it works.

Logistic Regression is used to estimate the probability that a sample datum belongs to a class - such as whether a loan applicant is more likely to repay or default on their loan. If the estimated probability is greater than some threshold, such as 0.5 then the sample belongs to the positive class - usually labelled as ‘1’. If less than the chosen threshold then it would belong to the negative class with a label of ‘0’.

However it enforces linear decision boundaries and can find it difficult to obtain complex relationships. For more challenging classification tasks with non linear decision boundaries more more robust results can be obtained using more complex solutions such as support vector machines and neural networks. Though these methods take more time to train and have more complex sets of hyper parameters.

Nevertheless Logistic Regression is a great for simple classification problems and small datasets. In this video we have looked at binary choices, but it can easily be extended to deal with multi-class, multinomial regression.

0:00 Introduction
0:14 MACHINE LEARNING
3:07 SIGMOID FUNCTION
6:33 ADDING A SECOND PARAMETER
8:00 TRAINING OUR ALGORITHM
10:28 LOGISTIC REGRESSION

#ai #artificialintelligence #machinelearning #logisticregression