Scribd
Upload
14 views3 pages

Red Wine Quality Dataset Class Imbalance Analysis

The assignment involves loading the Red Wine Quality dataset into WEKA Explorer to analyze class imbalance in the quality attribute. Techniques like Resample were applied to balance the classes, resulting in improved accuracy from 70.61% to 89.35%, along with enhanced precision and recall, particularly for minority classes. The findings conclude that resampling effectively mitigates class imbalance, leading to better overall classifier performance.

Uploaded by

mahatonikita00
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
14 views3 pages

Red Wine Quality Dataset Class Imbalance Analysis

The assignment involves loading the Red Wine Quality dataset into WEKA Explorer to analyze class imbalance in the quality attribute. Techniques like Resample were applied to balance the classes, resulting in improved accuracy from 70.61% to 89.35%, along with enhanced precision and recall, particularly for minority classes. The findings conclude that resampling effectively mitigates class imbalance, leading to better overall classifier performance.

Uploaded by

mahatonikita00
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

ASSIGNMENT 7

Problem Statement: Load the Red Wine Quality dataset into WEKA Explorer and examine
it under the Preprocess tab. Check for class imbalance in the quality attribute. If imbalance
exists, apply techniques like Resample or SMOTE to balance the classes. Then, use a classifier
under the Classify tab with 10-fold cross-validation. Compare performance (accuracy,
precision, recall) before and after handling imbalance. Report your findings and comment on
the effect of balancing the data.

Solution:

1. Loading and Analyzing the Red Wine Quality Dataset in WEKA Explorer
 The dataset is in ARFF format and consists of 11 numeric attributes (V1 to V11)
and a nominal class attribute Class with values {1,2,3,4,5,6} representing wine
quality.

 Class Distribution Imbalance:

 The dataset’s class attribute (Class) is categorical with values {1, 2, 3, 4, 5, 6},
representing different quality levels of wine.

 Classes 3 and 4 are the majority classes, while 1, 2, 5, and 6 are underrepresented,
indicating an imbalance in the class distribution.

2. Applying Resample Technique in WEKA to Address Class Imbalance

 Filter applied: Supervised > Instance > Resample

1
 Parameter modified: BiasToUniformClass = 1.0

3. Classifier Evaluation on Imbalanced Dataset (Before Applying Resample)

 Selected RandomForest from: Classify > Choose > trees > RandomForest
 Set to 10-fold cross-validation to ensure a robust performance estimate.

2
4. Classifier Evaluation on Balanced Dataset (After Applying Resample)

 Same classifier (RandomForest) and 10-fold cross-validation method used for


consistency.

5. Findings and Effect of Balancing the Data Using Resample

1. Accuracy Improved:
o Classifier accuracy increased from 70.61 % (before) to 89.35% (after
Resample).
2. Enhanced Precision and Recall:
o Precision and recall scores improved, especially for minority classes.
3. Reduced Class Bias:
o Model predictions became less biased toward majority classes.
4. Better Class Coverage:
o After resampling, the classifier showed more balanced performance across
all classes.
5. Conclusion:
o Resampling effectively handled class imbalance, leading to improved
overall classifier performance and fairer results.

You might also like