📘

Bagging Aggregation

Implement bagging aggregation for an ensemble of model predictions.

Given predictions from $m$ models for $n$ data points:

For classification (mode="classify"): return the majority vote for each data point
For regression (mode="regress"): return the mean prediction for each data point

In case of a tie in majority voting, return the smallest class label.

Return a list of aggregated predictions, rounded to 4 decimal places for regression.

Input:

predictions = [[1, 0, 1], [0, 0, 1], [1, 1, 1]]
mode = "classify"

Output:

[1, 0, 1]

Reasoning:

The input predictions is a 2D list where each row represents a model's predictions and each column represents a data point.
For each data point (column), we count the occurrences of each class label:
- For the 1st data point, the counts are 2 (for class 1) and 1 (for class 0).
- For the 2nd data point, the counts are 1 (for class 1) and 2 (for class 0).
- For the 3rd data point, the counts are 3 (for class 1) and 0 (for class 0).
We apply the majority vote for each data point:
- The 1st data point has a tie, but since 0 is the smallest class label in case of a tie, it's not selected; instead, the smallest label among the tied ones is chosen which is $1$ .
- The 2nd data point has a majority vote for class $0$ .
- The 3rd data point has a majority vote for class $1$ .
The final output is the list of majority votes for each data point: $[1, 0, 1]$

Editor

Python 3.13.1

0/0

Run code to see test results.