I’m so confused about the differences between statistical analysis and machine learning. When I make a linear regression model in machine learning, the main purpose is to predict the target value regarding to a new feature. And also, in case of statisical analysis, I did same thing.
3 Likes
This is a bit complex indeed. I would use the terminology used in this book (Chapter 2). Everything that you mentioned could be part of “statistical learning”. Under this term, there is 2 paradigms:
- prediction: you are interested to learn the model that provides the best possible predictions. We refer to machine learning here such that we learn a predictive model.
- inference: you are interested to learn a model such that you can understand the relationship between the data and the target. We refer to statistical analysis since you want to check the relation between the data and target.
Therefore, the main difference between the two approaches is the end goal for which you want to use your model: prediction vs. analysis.
NB: Sometimes, we learn a predictive model and then try to understand as well the mechanism that provides these predictions (this is called interpretation, inspection, or explanability).
NB2: In addition, you have descriptive statistics that compute statistics describing the data.
12 Likes
Thank you so much!!