Question

We have three candidate predictive models shown below automatically generated to solve the regression task of predicting variable y given a another variable x. All three models are approximations of the underlying function given only the provided training data. Recall that supervised data mining is difficult due to the issue of picking one of the infinite number of potential predictive models. The true underlying relationship is denoted by the green function Predictive Model 1 Predictive Model 2 Predictrve Model 3 -Model Model True function Model True function Samples True function .Samples . Samples a. Out of the three models, which one would provide the most accurate predictions in the real world? Justify your answer b. Which of the three models obtains the highest performance on the training data? Secondly, discuss the real-world performance of this predictive model. Justify your answers. c. Which of the three models overfits on the data? Justify your answer. d. Which of the three models underfits on the data? Justify your answer

Answer 1