Train and test a simple regression model.
-
-
- Train and test a simple regression model. SIMPLE means that your model with include only ONE explanatory variable. Choose the feature you believe is the best one for this model. (5 points)
-
-
-
-
- What is the MAPE for your training data?
-
-
-
-
-
- What is the MAPE for your test data?
-
-
-
-
-
- Is there any evidence of over-fitting? Explain your answer.
-
-
-
-
-
- In words, how would you describe the parameter estimate of the single feature in your model?
- Train and test a second regression model – this model should include ONE continuous numerical feature and ONE categorical variable (note: you may have multiple indicator variables representing a single categorical variable). You can include the indicator variables as either intercept shifters, slope-shifters, or both (you decide). (15 points)
- What is the MAPE for your training data?
- What is the MAPE for your test data?
- Is there any evidence of over-fitting? Explain your answer.
- In words, interpret TWO of the estimated model parameters.
- Train and test a final regression model – this model should include as many features as you want. You decide what this final set of predictors should be. Feel free to explore transformations of each feature, e.g. log transformations, power transformations, interactions/combinations of features, etc. (20 points)
- What is the MAPE for your training data?
- What is the MAPE for your test data?
- Is there any evidence of over-fitting? Explain your answer.
- Conclusion – answer the following questions? (10 points)
-
-
-
-
-
- Are you satisfied with your final model’s performance? Why or why not?
- What do you propose for next steps?
- What would you do differently if you could?
- Based on your findings – what would you want to communicate to business leadership in this industry?
-
-