A practical introduction to data preprocessing

Abstract

When the input data of a machine learning model undergoes the proper preprocessing, it can drastically affect the results. In Ben Gurion University’s paper about the HHD dataset, they discuss how their CNN model achieved 72.57% accuracy. When the same dataset was used to train a CNN after using the preprocessing steps mentioned above, the model achieved 84% accuracy. While Ben Gurion University does not discuss the preprocessing used in their model, it is clear that proper preprocessing of the data will play a considerable role in the success of the model. (from Conclusion)