---
kind: concept
slug: synthetic-minority-oversampling
title: Synthetic minority oversampling (SMOTE)
summary: >-
  SMOTE adds synthetic examples of a rare class so a model does not learn only the majority case. Two cards use it on tabular data. One reports a LightGBM model without any no-SMOTE run. The other, one run on one dataset, shows SMOTE raising recall from 0 to 0.59 while accuracy falls to 0.58. Both rest on single author-reported results.
tags: [class-imbalance, data-augmentation, tabular]
papers: ["2108.00071", "2310.00505"]
related: [gradient-boosted-decision-trees, class-balanced-loss, data-augmentation, data-resampling]
status: draft
updated_at: 2026-11-06
---

## Overview

SMOTE is an oversampling method that makes a rare class more common by adding
synthetic examples. It creates each new minority point by interpolating between
a minority instance and nearby minority instances in feature space
[arxiv:2108.00071 §4.2.2]. The exact interpolation details are left to the
original SMOTE paper, so this wiki does not state them.

Two cards use SMOTE on tabular classification data, in different roles. One
applies it once before a gradient-boosted model [arxiv:2310.00505 §3.4]. The
other compares it with other resampling methods under logistic regression
[arxiv:2108.00071 §4]. Neither states an invariant the other shares, so no
shared invariant is listed.

## Variants

### SMOTE as one step before LightGBM

The card is [Unveiling the Unborn: Advancing Fetal Health Classification
through Machine Learning](/papers/2310.00505). The pipeline takes 22 numeric
features from cardiotocography records, applies SMOTE to balance the classes
and fits a LightGBM classifier [arxiv:2310.00505 §3.2] [arxiv:2310.00505 §3.3]
[arxiv:2310.00505 §3.4]. The model settings are listed on the card
[arxiv:2310.00505 §3.4]. The SMOTE parameters (neighbour count, sampling
strategy) and the point in the pipeline where SMOTE runs are not stated
[arxiv:2310.00505 §3.4] [arxiv:2310.00505 §4.4]. The paper reports no run
without SMOTE, so its effect on the reported scores is not measured
[arxiv:2310.00505 §5].

### SMOTE compared with other resamplers under logistic regression

The card is [Foundations of data imbalance and solutions for a data
democracy](/papers/2108.00071). It is a tutorial chapter that fits logistic
regression on the Porto Seguro Safe Driver Prediction data (about 595k rows,
22 of 59 columns kept) [arxiv:2108.00071 §3]. It compares SMOTE with random
oversampling, ADASYN, random undersampling, Tomek links and edited nearest
neighbours, against a baseline with no resampling [arxiv:2108.00071 §4]. The
SMOTE run uses five neighbours and seed 42 [arxiv:2108.00071 §4.2.2].

Author-reported, single run [arxiv:2108.00071 §4]:

- No resampling: accuracy 0.96, recall 0, AUC 0.62. The baseline finds no
  positive cases.
- SMOTE: accuracy 0.58, precision 0.57, recall 0.59, F1 0.58, AUC 0.62.
- Random oversampling: accuracy 0.59, recall 0.60, AUC 0.63.
- ADASYN: accuracy 0.58, recall 0.59, AUC 0.62.

The chapter concludes that oversampling methods do better than undersampling
here [arxiv:2108.00071 §5]. The gaps among the three oversamplers are 0.01 in
accuracy or AUC, and the chapter does not test them [arxiv:2108.00071 §4].

## When to use which

| Situation | Use | Why |
|---|---|---|
| Rare class in tabular data, tree-based model | SMOTE before LightGBM, as in the first variant | This is the only tree-based use in the wiki [arxiv:2310.00505 §3.4]. Its result is author-reported on one dataset with no baseline [arxiv:2310.00505 §5]. |
| Logistic regression, and finding rare cases matters more than accuracy | SMOTE, expecting lower accuracy | In one run SMOTE raised recall from 0 to 0.59 and lowered accuracy from 0.96 to 0.58 [arxiv:2108.00071 §4]. The baseline never predicts the rare class. |
| Choosing between SMOTE, random oversampling and ADASYN | No choice is supported by the wiki | The three differ by about 0.01 in accuracy and AUC on one run, and the chapter does not test these gaps [arxiv:2108.00071 §4]. |
| Need to show that oversampling helps | Run your own comparison with and without SMOTE, split first | The first card has no run without SMOTE [arxiv:2310.00505 §5]. The second has a baseline but a single run [arxiv:2108.00071 §4]. |
| Cross-validation or a held-out split | Oversample the training part only | If test records enter the oversampling, the scores can be inflated. The first card does not state its order [arxiv:2310.00505 §3.4] [arxiv:2310.00505 §4.4]. The second appears to resample before splitting (see Contradictions) [arxiv:2108.00071 §4]. |

## Implementation sketch

The sketch uses the split-first order. The first card does not state its
order, so that step is our recommendation. Steps the papers leave open are
marked.

```pseudocode
# Split first, then oversample the training part only
# Order in arxiv:2310.00505 not stated  [arxiv:2310.00505 §3.4]
# Order in arxiv:2108.00071 appears to be the reverse  [arxiv:2108.00071 §4]
def fit_with_smote(features, labels):
    # 80/20 split as in the first card  [arxiv:2310.00505 §3.4]
    # Shuffling and stratification not stated
    train_x, train_y, test_x, test_y = split(features, labels,
                                             test_size=0.2)

    # Five neighbours and seed 42 as in the second card  [arxiv:2108.00071 §4.2.2]
    # The first card does not state these  [arxiv:2310.00505 §3.4]
    res_x, res_y = smote(train_x, train_y, k_neighbors=5,
                         random_state=42)

    # LightGBM settings as stated in the first card  [arxiv:2310.00505 §3.4]
    model = LGBMClassifier(learning_rate=0.1, n_estimators=100,
                           num_leaves=31, random_state=123)
    model.fit(res_x, res_y)
    return model, test_x, test_y
```

## Contradictions

- **Accuracy values in the first card.** The card reports 98.31% in the
  abstract and 98.32% in §5 [arxiv:2310.00505 §5]. Its model comparison gives
  97.91% for LightGBM [arxiv:2310.00505 §5]. The paper does not say which
  split or run produces each figure, so these scores cannot be attributed to
  SMOTE on their own.
- **SMOTE's effect on accuracy across the two cards.** The first card reports
  about 98% accuracy with SMOTE [arxiv:2310.00505 §5]. The second reports
  accuracy falling from 0.96 to 0.58 when SMOTE is used [arxiv:2108.00071 §4].
  The datasets, models, splits and metrics differ, and the first card has no
  run without SMOTE. The two results are not comparable, so neither confirms
  or refutes the other.
- **Precision and recall in the second card.** The card's own confusion
  matrices do not match the reported precision and recall in several tables.
  For the SMOTE table, the reported values appear swapped relative to the
  matrix, and the card attributes this to the argument order in the chapter's
  code [arxiv:2108.00071 §4]. Check these before quoting the SMOTE row.
- **Resampling order in the second card.** The chapter's code appears to call
  resampling on the full data before the split. The card infers this from
  the size of the test sets, and the chapter does not state the order
  [arxiv:2108.00071 §4]. If so, the reported SMOTE scores may be inflated
  [arxiv:2108.00071 §4].

## Papers

- [Foundations of data imbalance and solutions for a data democracy](/papers/2108.00071): tutorial comparison of SMOTE, random oversampling, ADASYN and undersampling with logistic regression on one dataset, one run per method.
- [Unveiling the Unborn: Advancing Fetal Health Classification through Machine Learning](/papers/2310.00505): applies SMOTE to balance the classes before a LightGBM classifier on 22 cardiotocography features; no run without SMOTE is reported.