There are multiple reasons to use data to get detailed insights into what is going on. That is what the machine learning process is all about. However, you eventually get to a model. Inference is when you use that model to learn about things that happened in the past. However, prediction is when you use your model to make predictions in the future. These two ways of working with models mean that you need to understand them clearly.
Predictive modeling and generative modeling are two of the many things you can do with machine learning models.
Understanding the Past
Model interpretability is one of the cornerstones of inference. After all, how can you understand the past if you don’t actually have a good idea of what went into generating the answers you have. Because of this, it is crucial for you to understand what happens with machine learning models after you have generated them.Generative modeling is at the heart of what it means to do inference the right way. When all of these things come together, it typically leads to much better results for your understanding of how machine learning and data science work.
The answers you get will ultimately come down to how much you understand the inner workings of your model. It is one of the many reasons why there are only a few models that can be interpreted by humans. Choosing how you work with machine learning models is, therefore, a crucial component of this process as well.
Predicting the Future
Predictive modeling does the opposite. This type of modeling takes your machine learning models and uses them to look forward to the future. These models are excellent for people who just want a good idea of what is going on. For example, if you had historical data and you wanted to look at what things could be in the future, you would use predictive modeling to accomplish that task. Model interpretability doesn’t become as big of an issue because of how things work when you do it this way. The process you use for prediction and inference is also slightly different.
Understanding the Workflows
The most important thing is that predictive modeling and generative modeling are similar enough that there are commonalities to take away. However, they are also different enough that the two communities can come together and learn things from each other. For example, generative modeling is typically what people in statistics do. Predictive modeling is what machine learning is all about.
As a practitioner, it is crucial to understand other ways of doing things, as it will help you do your job better. It might also help you find blind spots in your thinking that you did not see before. Because of this reality, there will always be something for you to understand. You might also be faced with situations that require you to understand which type of modeling is required for your current project.