What or Why in Machine Learning
A comprehensive guide to interpreting models using Python Machine learning using big data is all the rage in business. These phrases are giving “synergistic office speak” a run for their money. Behind these buzzwords, the development of machine learning techniques and the machines that implement them in the past decade has been truly remarkable. The increasing complexity of models has allowed for machines to better classify, label, and predict continuous values. However, as the models become more complex, how can we be sure the models are not utilizing training biases or predicting on subtle changes to the background noise. Machines make errors differently than humans. (See examples here , here , and here ) Using the python libraries ELI5 , PDPbox , Lime , and SHAP , we can visualize how a model predicts an outcome, weights the importance of features, or distinguishes boundaries in an image. Without further ado, let’s peek behind the curtain of the black box model to see how our mod