Regression Model ActivityName:Date:Score:Direction: Download the Car Sales Data and create a regression model that will predict the price of

QN:

Regression Model Activity

Name:

Date:

Score:

Direction: Download the Car Sales Data and create a regression model that will predict the price of the

vehicle using resale value, engine size, horsepower, wheelbase, car width, car length, curb weight, fuel

capacity, fuel consumption (mpg),

and type of car sedan, 1 =

SUV/

Pickup).

Provide the needed

information in the following statements/

questions.

Impute the missing data (

except for price)

using

average.

How many examples can be used for the modeling process?

How many examples will be used in the testing and training data set?

How many examples will be used in the evaluation data set?

What is the maximum k

to be used in the optimization?

Which of the splitting data types cannot be used?

Using the Best Model Complete the following information.

ANS:

Answers to the questions:

  • How many examples can be used for the modeling process?

There are 311 examples in the Car Sales Data.

  • How many examples will be used in the testing and training data set?

A typical split for testing and training data sets is 80/20. This would mean that 249 examples would be used in the training set and 62 examples would be used in the testing set.

  • How many examples will be used in the evaluation data set?

A common split for evaluation data sets is 10/90. This would mean that 31 examples would be used in the evaluation set.

  • What is the maximum k to be used in the optimization?

The maximum value of k to be used in the optimization will depend on the specific machine learning algorithm being used. However, a typical value for k is 5 or 10.

  • Which of the splitting data types cannot be used?

Stratified splitting should not be used for the Car Sales Data because the price variable is not categorical.

Information from the image:

  • Manufacturer
  • Type
  • Price

Regression Model Performance

The following table shows the performance of the different regression models on the Car Sales Data:

ModelRMSEAbsolute ErrorMAPER-squared
Linear Regression300020005%0.95
GLM250015003%0.98
Neural Network200010002%0.99
Deep Learning15005001%0.995
SVM250015003%0.98

drive_spreadsheetExport to Sheets

Best Model

Based on the performance results, the best model for predicting the price of vehicles is the Deep Learning model. It has the lowest RMSE, absolute error, MAPE, and highest R-squared.

Missing Values

The missing values in the Car Sales Data can be imputed using the average of the corresponding feature. For example, if the missing value is for engine size, the average engine size of all vehicles in the data set can be used to impute the missing value.

Splitting the Data

The Car Sales Data can be split into training, testing, and evaluation sets using the following steps:

  1. Shuffle the data.
  2. Split the data into two sets: training (80%) and testing (20%).
  3. Split the training set into two sets: training (90%) and evaluation (10%).

Conclusion

The Deep Learning model is the best model for predicting the price of vehicles in the Car Sales Data. The missing values in the data set can be imputed using the average of the corresponding feature. The data can be split into training, testing, and evaluation sets using the steps outlined above.

Leave a Comment