Predicting White Wine Quality Scores RAPHAEL MWANGI.
-
Upload
martin-chase -
Category
Documents
-
view
214 -
download
2
Transcript of Predicting White Wine Quality Scores RAPHAEL MWANGI.
![Page 1: Predicting White Wine Quality Scores RAPHAEL MWANGI.](https://reader036.fdocuments.in/reader036/viewer/2022080222/56649ce15503460f949ac82a/html5/thumbnails/1.jpg)
Predicting White Wine Quality ScoresRAPHAEL MWANGI
![Page 2: Predicting White Wine Quality Scores RAPHAEL MWANGI.](https://reader036.fdocuments.in/reader036/viewer/2022080222/56649ce15503460f949ac82a/html5/thumbnails/2.jpg)
Background
Tree-models are intuitive, easy to understand, and can be used as tools for more advanced algorithms to create highly accurate models
Tree-modeling requires practically no knowledge of statistics to understand
![Page 3: Predicting White Wine Quality Scores RAPHAEL MWANGI.](https://reader036.fdocuments.in/reader036/viewer/2022080222/56649ce15503460f949ac82a/html5/thumbnails/3.jpg)
Basic Example
Terminology:o “Node” – denoted ,a circle in which observations
are grouped into by a particular If Then statement
o “” = the number of observations in the node “”o = the average wine quality score of the N
observations in the node “”.
If Alcohol < 10.85
If Alcohol > 10.85
=3898=5.87
=2465=5.596
=1433=6.345
𝑹𝟏
𝑹𝟐 𝑹𝟑
![Page 4: Predicting White Wine Quality Scores RAPHAEL MWANGI.](https://reader036.fdocuments.in/reader036/viewer/2022080222/56649ce15503460f949ac82a/html5/thumbnails/4.jpg)
Actual Tree Model
![Page 5: Predicting White Wine Quality Scores RAPHAEL MWANGI.](https://reader036.fdocuments.in/reader036/viewer/2022080222/56649ce15503460f949ac82a/html5/thumbnails/5.jpg)
Bagging Bagging
Take a large number of samples, “B”, of size n from your dataset, with replacement (bootstrap samples)
Fit a tree model to each bootstrap sample. When making predictions of y for specific values
of x, average from your “B” bootstrap samples for a more accurate prediction
![Page 6: Predicting White Wine Quality Scores RAPHAEL MWANGI.](https://reader036.fdocuments.in/reader036/viewer/2022080222/56649ce15503460f949ac82a/html5/thumbnails/6.jpg)
Results of Single Tree Model vs. Bagged Tree Model
Each prediction we make will, on average, be:o 0.77 points off from the true wine
quality score for the Single Tree Model
o 0.75 points off from the true wine quality score for the Bagged Tree Model
Bagged model did slightly better