Interactions between predictive analytics, ML, and case-based and rules-based reasoning

Question

For my studies on economy, I work on prediction of judicial decisions. I don't really understand the interactions between several concepts:

• predictive analytics
• machine learning
• case-based reasoning
• rules-based reasoning

At the moment, I understand that predictive analytics is an area of statistics that deals with extracting information from data and using it to predict a outcome, like a judicial decision. In predictive analytics there are 2 kinds of tools: econometric tools like MCO regression and machine learning tools like neural networks.

Among ML tools, we can distinguish between 2 approaches: rules-based reasoning and case-based reasoning. Given a given legal problem, with the rules approach, the algorithm learns the rules that led to these data, deduce a linear model and then we can confront the problem to the model to deduce the solution given by a court. With cases approach, the algorithm brings the given problem closer to similar cases found in the data, and deduces a solution given by a court.

Is it correct?

Are different kind of ML tools like K-NN or neural network only used for one type of reasoning? or they can be used for both?

Answer 1

Personally, I'm not enamored of the "rules-based reasoning vs case-based reasoning" categorization; I'm not sure it is a particularly useful mindset for modern machine learning. For instance, deep learning is one of the most effective ML methods used today in several domains, yet it doesn't seem to fall into either of those categories. That's just my opinion.

Rather than trying to find this kind of high-level generalization about how ML works, I think it might be more useful to study some specific ML techniques: linear regression, k-nearest neighbors, SVM, random forest, deep learning, etc.