how details, results, and answers to questions. Our goal is to modify our predictor variables to get the best model possible. Link to data set to use : https://archive.ics.uci.edu/ml/datasets/YearPredictionMSD The first column is the year; this is the target variable. Part I Transform the predictors: For each of the 90 predictors, determine which transformation(s) is (are) best. * no transformation * log transform * exponential transform * inverse transform * square transform * square root transform * center and scale
Use R .
Please do not take this problem if you can not use the data provided.
Show details, results, and answers to questions.
Our goal is to modify our predictor variables to get the best model possible.
Link to data set to use :
https://archive.ics.uci.edu/ml/datasets/YearPredictionMSD
The first column is the year; this is the target variable.
Part I Transform the predictors:
For each of the 90 predictors, determine which transformation(s) is (are) best.
* no transformation
* log transform
* exponential transform
* inverse transform
* square transform
* square root transform
* center and scale
* range transform
* interval binning
* quantile binning
Make a table listing the predictor 1-90, and the transform you selected. Make a summary table showing how many times each transform was used.
Part II Principal Component Analysis:
Compute the principal components of the original data.
Use visualizations and PCA statistics to determine how many components you should use.
Part III Regression Models:
Build a model that predicts the song's year with your untransformed predictors. Print your model's test set root mean squared error in the first line of your document after the title.
Build a model that predicts the song's year with your transformed predictors. Print your model's test set root mean squared error in the first line of your document after the title.
Build a model that predicts the song's year with the principal components. Print your model's test set root mean squared error in the first line of your document after the title.
Untransformed model RMSE: XXX
Transformed model RMSE: XXX
Principal component model RMSE: XXX
train set : first 463715 rows
test set: last 51630 rows
You may use any supervised model structure for your predictive models, but all three must have the same structure.
Trending now
This is a popular solution!
Step by step
Solved in 4 steps with 2 images
