# What benefits do using log returns for model training provide?

I came across a paper that uses Support Vector Machines to classify a buy/sell/hold decision each hour at the $$\pm$$0.5% threshold. The paper can bee seen here. The paper yielded impressive predictive power as well as high returns. It was noted that during the training phase they created the [1, 0, -1] labels by computing the percentage rate of change on log hourly returns.

I have looked at answers provided here Why should we use log returns? Log normality, however this wasn't in the context of label creation for an ML classification problem.

I was wondering if this technique offers predictive insight that using normal percentage returns? And if so, why does this work?

• Just idea, maybe the feature $\frac{x-y}{y} \approx \ln(\frac{x}{y}) = \ln{x}-\ln{y}$ for $x$ close to $y$ can make something more simply? May 3, 2020 at 21:11
• But what benefit would that yield over using just returns without the log? May 3, 2020 at 21:25
• Sorry, I do not have time to go through whole article but I know that using a logarithm of some value can make other calculations more simple. Moreover, logarithm transform multiplication to addition which has lower computational cost..But as I said, just ideas. May 3, 2020 at 21:30