There seems to be a mistake here!

[PenaBrasil]

I have been developing and testing models for the financial market for some time. An accuracy of more than 90% seems impossible to me for such simple graphic samples (with little data). I replicated your model and got to 53%. By the way, the same thing was described in the "RandomForest Classifier results" issue. I began to doubt the results of this work. Can you explain these results?

[MasterBeard]

I also took the time to read their article because I noticed it had a lot of citations on Google Scholar. However, I saw that their article included a comparison result with Patel. Having read Patel's article published in Expert System, I know that their study does not divide the training and testing sets according to the time series. So, when I saw this comparison, I was puzzled. How can a model that divides data according to the time sequence be compared with one that does not?