I mange virkelige datasett oppstår det ofte et problem med skjev klassefordeling—der én klasse (majoriteten) har langt flere forekomster enn den andre (minoriteten). Denne skjevheten kan føre til at modeller favoriserer majoritetsklassen, noe som reduserer treffsikkerheten for minoritetsklassen. En vanlig løsning er **undersampling**, hvor antallet forekomster i majoritetsklassen tilfeldig reduseres slik at det samsvarer med antallet i minoritetsklassen. Denne oppgaven gir praktisk erfaring med denne teknikken. Du får et DataFrame som inneholder en kategorisk målkolonne med to klasser. Målet er å returnere et nytt DataFrame der begge klassene er representert med likt antall forekomster, oppnådd ved å tilfeldig undersample majoritetsklassen.


import unittest
import user_code
import ast
import re   
import importlib
import csv
import unittest
import pandas as pd
import importlib
import sys

class TestTask(unittest.TestCase):
    def setUp(self):
        if "user_code" in sys.modules:
            del sys.modules["user_code"]
        import user_code
        importlib.reload(user_code)
        self.undersample_majority_class = getattr(user_code, 'undersample_majority_class', None)

    def test_balanced_result(self):
        import user_code
        data = {
            "feature": list(range(12)),
            "target":  ["A"] * 8 + ["B"] * 4
        }
        df = pd.DataFrame(data)
        balanced = self.undersample_majority_class(df, "target") if self.undersample_majority_class else None
        _dynamic_test(
            self,
            balanced is not None and set(balanced['target'].value_counts()) == {4},
            "Returned DataFrame contains both classes in equal numbers",
            f"Expected both classes to have 4 samples, got counts: {None if balanced is None else balanced['target'].value_counts().to_dict()}"
        )

    def test_minority_preserved(self):
        import user_code
        data = {
            "feature": list(range(8)),
            "target":  ["A"] * 2 + ["B"] * 6
        }
        df = pd.DataFrame(data)
        balanced = self.undersample_majority_class(df, "target") if self.undersample_majority_class else None
        minority_rows = df[df["target"] == "A"]
        if balanced is not None:
            for idx, row in minority_rows.iterrows():
                match = ((balanced["feature"] == row["feature"]) & (balanced["target"] == row["target"]))
                _dynamic_test(
                    self,
                    match.any(),
                    "All original minority class samples are preserved",
                    f"Minority sample {row.to_dict()} not found in balanced DataFrame"
                )
        else:
            _dynamic_test(self, False, "", "Function returned None")

    def test_majority_count(self):
        import user_code
        data = {
            "feature": list(range(10)),
            "target":  ["X"] * 7 + ["Y"] * 3
        }
        df = pd.DataFrame(data)
        balanced = self.undersample_majority_class(df, "target") if self.undersample_majority_class else None
        if balanced is not None:
            counts = balanced['target'].value_counts()
            _dynamic_test(
                self,
                counts["X"] == counts["Y"] == 3,
                "Majority class is undersampled to match minority class count",
                f"Expected 3 samples for each class, got: {counts.to_dict()}"
            )
        else:
            _dynamic_test(self, False, "", "Function returned None")

    def test_already_balanced(self):
        import user_code
        data = {
            "feature": list(range(6)),
            "target":  ["A", "A", "A", "B", "B", "B"]
        }
        df = pd.DataFrame(data)
        balanced = self.undersample_majority_class(df, "target") if self.undersample_majority_class else None
        if balanced is not None:
            counts = balanced['target'].value_counts()
            _dynamic_test(
                self,
                counts["A"] == counts["B"] == 3 and len(balanced) == 6,
                "Function works when DataFrame is already balanced",
                f"Expected 3 samples for each class, got: {counts.to_dict()}"
            )
        else:
            _dynamic_test(self, False, "", "Function returned None")

    def test_different_labels(self):
        import user_code
        data = {
            "feature": list(range(9)),
            "target":  [1, 1, 1, 1, 0, 0, 0, 0, 0]
        }
        df = pd.DataFrame(data)
        balanced = self.undersample_majority_class(df, "target") if self.undersample_majority_class else None
        if balanced is not None:
            counts = balanced['target'].value_counts()
            _dynamic_test(
                self,
                set(counts) == {4},
                "Function works for any two-class DataFrame, regardless of class labels",
                f"Expected both classes to have 4 samples, got: {counts.to_dict()}"
            )
        else:
            _dynamic_test(self, False, "", "Function returned None")

def _dynamic_test(test_case, condition, success_message, failure_message):
    if condition:
        test_case._testMethodName = success_message
        test_case.assertTrue(True, success_message)
    else:
        test_case._testMethodName = failure_message
        test_case.fail(failure_message)

def normalize_text(text):
    text = text.lower()
    text = re.sub(r"\\s{2,}", " ", text)
    text = re.sub(r"\\s*([,:?])\\s*", r"\\1 ", text)
    return text.strip()

def change_var(code: str, var_name: str, value: str) -> str:
    tree = ast.parse(code)
    lines = code.splitlines()
    changed = False
    # Collect all assignment nodes to modify
    assign_nodes = [
        (i, node)
        for i, node in enumerate(tree.body)
        if isinstance(node, ast.Assign)
        and any(isinstance(target, ast.Name) and target.id == var_name for target in node.targets)
    ]

    # If nothing to change, return unmodified code
    if not assign_nodes:
        return code

    # Perform replacements for all matching assignments (from last to first to not break line offsets)
    for i, node in reversed(assign_nodes):
        start_line = node.lineno - 1
        line = lines[start_line]
        indent = ' ' * (len(line) - len(line.lstrip()))
        lines[start_line] = f"{indent}{var_name} = {value}"
        next_line = len(lines)
        for next_node in tree.body[i+1:]:
            if hasattr(next_node, 'lineno'):
                next_line = next_node.lineno - 1
                break
        if next_line > start_line + 1:
            lines[start_line+1:next_line] = []
        changed = True

    return '\\n'.join(lines) if changed else code

if __name__ == "__main__":
    unittest.main()


test_main.py

Et praktisk, hands-on kurs for ambisiøse dataforskere som er klare til å møte utfordringer med store datasett i virkelige situasjoner. Lær å behandle, utvalge og analysere massive datasett effektivt ved hjelp av Python og essensielle biblioteker. Hver seksjon inneholder engasjerende videoforklaringer og interaktive utfordringer for å bygge din ekspertise.

Lær grunnleggende strategier for håndtering av datasett som er for store til å få plass i minnet, inkludert chunking- og streamingteknikker.

Utforsk metoder for å balansere og ta utvalg fra store datasett, inkludert oversampling og undersampling.

Lær å bruke polars-biblioteket for rask og minneeffektiv databehandling.

Utfordring: Bruk Undersampling

Løsning