Why might a data mining professional prefer a neural network over a decision tree model?

• A. They always require less data
• B. They are better at finding the strength of connections between attributes
• C. They train faster on small datasets
• D. They are easier to interpret

Answer: (B)

Neural networks encode the influence of inputs through connection weights learned during training, so they can capture not only individual effects but also how strongly attributes interact to affect the outcome. This ability to model and quantify the strength of relationships and interactions between attributes is a key strength, especially for complex, non-linear patterns that a single decision tree splitting on feature values might miss. A decision tree provides interpretable rules and local decisions, but it doesn’t directly expose or optimize the overall strength of connections between inputs. Of course, neural networks typically require more data and computation and are harder to interpret, which is why they aren’t always the best choice, but when the aim is to understand and leverage how strongly inputs influence the prediction, neural networks offer the best fit.