The workflow of any Machine Learning algorithm is simple. However, this workflow could become complex if training your data seemed to run forever! This happens when you have big data. The challenges of big data manifest in both inference and computation. However, you cannot do away with data. In analysis, there is never “sufficient” data. As you get more and more data, you can start subdividing the data to gain better insights. The only rescue to such problem is comparing the computational complexities of various Machine Learning algorithms that can provide a solution to your question and finding the best among them that can produce timely, accurate, and precise analysis and decision-making.
This article describes the computational complexities of Logistic Regression. Logistic Regression is used for binary classification in linearly separable data or almost linearly separable data. Before diving deep into the computational complexities of Logistic Regression, I would recommend you go through my blog on Train/Test Complexity and Space Complexity of Linear Regression. This blog describes Time and Space Complexity and the concepts of Big O from scratch. You will also get to know about the computational complexities of Linear Regression. So, if you are new to such concepts, you must have a look at my previous blog before you move on with this article.
Every machine learning algorithm works best under a given set of conditions. It’s important to ensure that your algorithm fits the assumptions / requirements for superior performance. You can’t use any algorithm in any condition. For e.g.: We can’t use linear regression on a categorical dependent variable. Because we won’t be appreciated for getting extremely low values of adjusted R² and F1 Score. Instead, in such situations, we should try using algorithms such as Logistic Regression, Decision Trees, Support Vector Machine (SVM), Random Forest, etc.
#supervised-learning #logistic-regression #classification #computational-complexity #big-o