paulpas Data Quality
Comprehensive guide to data quality in machine learning and data science workflows.
When to Use This Skill
- Solving real-world data collection & ingestion problems
- Building machine learning pipelines with data quality
- Implementing best practices for data quality
- Optimizing model performance using data quality techniques
- Learning industry-standard approaches to data quality
When NOT to Use This Skill
- When using pre-built libraries without understanding underlying concepts
- For toy problems that don't require data quality rigor
- When domain expertise in specific problem requires different approach
- If your problem doesn't require the complexity this skill provides
Purpose and Key Concepts
Data Quality is a critical component of the machine learning workflow. This skill covers:
- Theoretical foundations — Mathematical principles and statistical concepts
- Practical implementation — Working code examples and patterns
- Common pitfalls — Mistakes to avoid and how to recover from them
- Best practices — Industry-standard approaches and optimization techniques
Core Workflow
- Understand the problem — Clearly define what you're solving for
- Select approach — Choose the right technique for your data and constraints
- Implement solution — Write clean, tested code following best practices
- Validate results — Verify your implementation with tests and validation
- Optimize performance — Improve efficiency and accuracy incrementally
Implementation Patterns
Pattern 1: Basic Data Quality
import pandas as pd
import numpy as np
from typing import Dict, Any
def basic_data_quality_check(df: pd.DataFrame) -> Dict[str, Any]:
"""
Perform basic data quality checks on a DataFrame.
Checks for missing values, duplicates, data types, and basic statistics.
"""
if not isinstance(df, pd.DataFrame):
raise TypeError("Input must be a pandas DataFrame")
quality_report = {
"total_rows": len(df)
"total_columns": len(df.columns)
"missing_values": df.isnull().sum().to_dict()
"duplicate_rows": int(df.duplicated().sum())
"data_types": df.dtypes.astype(str).to_dict()
"numeric_summary": {}
}
for col in df.select_dtypes(include=[np.number]).columns:
quality_report["numeric_summary"][col] = {
"mean": float(df[col].mean())
"std": float(df[col].std())
"min": float(df[col].min())
"max": float(df[col].max())
"null_count": int(df[col].isnull().sum())
}
return quality_report
Pattern 2: Production-Ready Data Quality
import logging
import pandas as pd
import numpy as np
from typing import Any, Dict, List, Optional
from sklearn.impute import SimpleImputer
from sklearn.preprocessing import StandardScaler
logger = logging.getLogger(__name__)
class ProductionDataQuality:
"""
Production-grade data quality handler following DRY principles.
Handles validation, imputation, scaling, and outlier detection.
"""
def __init__(self, missing_threshold: float = 0.5, outlier_std: float = 3.0):
self.missing_threshold = missing_threshold
self.outlier_std = outlier_std
self.imputer = SimpleImputer(strategy="median")
self.scaler = StandardScaler()
self.quality_issues: List[str] = []
def execute(self, data: pd.DataFrame) -> Dict[str, Any]:
"""Execute comprehensive data quality pipeline."""
if data.empty:
raise ValueError("Input DataFrame cannot be empty")
issues = []
clean_data = data.copy()
missing_cols = clean_data.columns[clean_data.isnull().mean() > self.missing_threshold]
if len(missing_cols) > 0:
issues.append(f"Dropping columns with >{self.missing_threshold*100}% missing: {list(missing_cols)}")
clean_data.drop(columns=missing_cols, inplace=True)
numeric_cols = clean_data.select_dtypes(include=[np.number]).columns
if len(numeric_cols) > 0:
clean_data[numeric_cols] = self.imputer.fit_transform(clean_data[numeric_cols])
for col in numeric_cols:
z_scores = np.abs((clean_data[col] - clean_data[col].mean()) / clean_data[col].std())
outlier_mask = z_scores > self.outlier_std
if outlier_mask.any():
issues.append(f"Detected {outlier_mask.sum()} outliers in column '{col}'")
clean_data.loc[outlier_mask, col] = clean_data[col].median()
if len(numeric_cols) > 0:
clean_data[numeric_cols] = self.scaler.fit_transform(clean_data[numeric_cols])
self.quality_issues = issues
return {
"cleaned_data": clean_data
"issues_found": issues
"rows_processed": len(data)
"columns_retained": len(clean_data.columns)
}
BAD vs GOOD Example
# BAD: Bypasses error handling, uses magic numbers, and mutates input silently
def bad_quality_check(df):
df.dropna()
df = df[(df['val'] > -999) & (df['val'] < 999)]
return df
# GOOD: Explicit validation, configurable thresholds, and proper return structure
def good_quality_check(df: pd.DataFrame, threshold: float = 0.5) -> pd.DataFrame:
if not isinstance(df, pd.DataFrame):
raise TypeError("Input must be a DataFrame")
if df.empty:
raise ValueError("DataFrame cannot be empty")
clean_df = df.dropna()
numeric_cols = clean_df.select_dtypes(include=[np.number]).columns
for col in numeric_cols:
q1, q3 = clean_df[col].quantile(0.25), clean_df[col].quantile(0.75)
iqr = q3 - q1
clean_df = clean_df[(clean_df[col] >= q1 - 1.5 * iqr) & (clean_df[col] <= q3 + 1.5 * iqr)]
return clean_df
Best Practices
- ✅ Always validate your implementation on test data
- ✅ Document your assumptions and methodology
- ✅ Use version control for reproducibility
- ✅ Monitor performance metrics in production
- ✅ Periodically review and update your approach
- ✅ Test with edge cases and outliers
- ✅ Log all significant operations for debugging
Common Pitfalls
| Pitfall | Problem | Solution | |
Constraints
MUST DO
- Validate all data preprocessing steps are fit-only on training data, never on validation or test sets
- Implement reproducible pipelines with fixed random seeds and deterministic operations where possible
- Report model performance with confidence intervals via bootstrapping or cross-validation across multiple runs
- Log all experiments with parameters, metrics, and artifacts using MLflow or equivalent tracking system
MUST NOT DO
- Do not evaluate a model on the same data used for training — always hold out a proper test set
- Avoid overfitting to the validation set by limiting hyperparameter search iterations
- Never use features that can only be computed at inference time (look-ahead bias)
- Do not report single-run accuracy without statistical significance testing or error bars
Live References
Authoritative documentation links for this skill's domain. The model follows markdown links at load time to resolve external references and inline content.