Machine Learning
Course Assignment
机器学习assignment代写 The assignment is to evaluate and describe a strategy based on predictions generated by a machine learning model.Part 1: Model···
The assignment is to evaluate and describe a strategy based on predictions generated by a machine learning model.
Part 1: Model Evaluation 机器学习assignment代写
Using the test predictions provided
- Generatethe Lift Chat and PR Curve for the attached predictions
- Togenerate the lift chart (using the Predictions vs Actuals Spreadsheet):
- Rank-orderthe data based on predictions (highest to lowest probability)
- Calculatethe cumulative sum of actuals for each row (e.g. at row 20 calculate the total sum of “1’s” in the actual column)
- Dividethe cumulative sum by the total number of positive examples
- Plotthe ordinal value of the sorted data on the x-axis, and the cumulative percentages on the y-axis
- Togenerate the PR Curve:
- Togenerate the lift chart (using the Predictions vs Actuals Spreadsheet):
- Rank-orderthe data based on predictions (highest to lowest probability)
- Calculateprecision as defined in the lectures
- Calculaterecall as defined in the lectures
- Plotthe PR-curve (precision vs recall)
- Referto the example spreadsheet as needed
Part 2: Next Steps 机器学习assignment代写
- As in the example, generate the profit curves (you can follow the example in the “Life Chart Exercise_example”spreadsheet to guide you, and use similar assumptions (e.g. expected profit / loss for a good / bad customer): the profit curve shows the expected profit that can be made from using the model to accept all customers below a certain prediction cut-off
- Estimatethe optimal probability cut-off to use when the model goes into
- Inone page or less, make a recommendation on how the strategy can be improved (e.g. by building other models beyond pure credit risk)
- For one of the additional models recommended, describe how the model can be integrated with businessstrategy (i.e. how a score will be converted into an action), how the model can be evaluated, and what potential risks or biases exist.