Imbalanced class handling (SMOTE, class weights) in ML Python - Model Pipeline Trace

Model Pipeline - Imbalanced class handling (SMOTE, class weights)

This pipeline shows how to handle imbalanced classes in a dataset using SMOTE to create synthetic samples and class weights to help the model learn better from minority classes.

Data Flow - 6 Stages

1Raw data input

1000 rows x 5 columns→Original dataset with imbalanced classes (90% class 0, 10% class 1)→1000 rows x 5 columns

[Feature1=2.3, Feature2=1.1, ..., Class=0]

↓

2Train/test split

1000 rows x 5 columns→Split data into training (80%) and testing (20%) sets→Train: 800 rows x 5 columns, Test: 200 rows x 5 columns

Train sample: [Feature1=2.3, ..., Class=0], Test sample: [Feature1=1.5, ..., Class=1]

↓

3SMOTE oversampling

Train: 800 rows x 5 columns→Create synthetic minority class samples to balance classes in training set→Train: 1440 rows x 5 columns (balanced classes)

Synthetic sample: [Feature1=2.0, Feature2=1.3, ..., Class=1]

↓

4Feature scaling

Train: 1440 rows x 5 columns, Test: 200 rows x 5 columns→Fit scaler on train features and scale (transform) both train and test features to similar ranges for model training→Train: 1440 rows x 5 columns (scaled), Test: 200 rows x 5 columns (scaled)

[Feature1=0.45, Feature2=0.33, ..., Class=1]

↓

5Model training with class weights

Train: 1440 rows x 5 columns (scaled)→Train model using class weights to emphasize minority class→Trained model

Class weights: {0:1.0, 1:1.0} after SMOTE (balanced)

↓

6Model evaluation

Test: 200 rows x 5 columns (scaled)→Evaluate model performance on imbalanced test set→Performance metrics

Accuracy=0.92, Precision=0.85, Recall=0.88

Training Trace - Epoch by Epoch

Loss
0.7 |*       
0.6 | *      
0.5 |  *     
0.4 |   *    
0.3 |    *   
0.2 |     *  
0.1 |       
    +--------
     1 2 3 4 5 Epochs

Epoch	Loss ↓	Accuracy ↑	Observation
1	0.65	0.70	Model starts learning, loss high, accuracy moderate
2	0.48	0.80	Loss decreases, accuracy improves
3	0.35	0.87	Model learns minority class better
4	0.28	0.90	Loss continues to decrease, accuracy rises
5	0.22	0.92	Training converges with good balance

Prediction Trace - 4 Layers

Layer 1: Input features

Layer 2: Hidden layer with ReLU activation

Layer 3: Output layer with sigmoid activation

Layer 4: Threshold decision

Model Quiz - 3 Questions

What is the main purpose of SMOTE in this pipeline?

A To scale features to the same range

B To remove samples from the majority class

C To create synthetic samples for the minority class

D To split data into train and test sets

Key Insight

Handling imbalanced classes with SMOTE and class weights helps the model learn minority classes better, improving overall performance and fairness.