Garbage In Garbage Out Everyone is aware of how important data is to the success of Supervised Machine Learning Projects. A lot of effort is spent on acquiring and cleansing the data. Then comes into play Feature generation. Project Teams often spend weeks generating the features before starting building the model. ML model is developed, tested, and evaluated in an iterative mode. In each iteration, features are regenerated and hyperparameters are fine-tuned to get a better model. If the model does not achieve the desired accuracy expected to make it a viable business solution, the project does not move forward because of a lack of support from business stakeholders. What options are left to improve the model? Can you evaluate the accuracy of labels used in the datasets? Can you hand-code the labels? Let’s look into various approaches to data labeling.