paulpas Metrics and KPIs
Comprehensive guide to metrics and kpis in machine learning and data science workflows.
When to Use This Skill
- Solving real-world experimentation & a/b testing problems
- Building machine learning pipelines with metrics and kpis
- Implementing best practices for metrics and kpis
- Optimizing model performance using metrics and kpis techniques
- Learning industry-standard approaches to metrics and kpis
When NOT to Use This Skill
- When using pre-built libraries without understanding underlying concepts
- For toy problems that don't require metrics and kpis 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
Metrics and KPIs 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 Metrics and KPIs
import pandas as pd
import numpy as np
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, mean_squared_error, r2_score
def compute_basic_metrics(y_true: pd.Series, y_pred: pd.Series, task_type: str = "classification") -> dict:
"""Compute basic evaluation metrics for classification or regression tasks."""
if y_true is None or y_pred is None:
raise ValueError("y_true and y_pred cannot be None")
if len(y_true) != len(y_pred):
raise ValueError("y_true and y_pred must have the same length")
metrics = {}
if task_type == "classification":
metrics["accuracy"] = accuracy_score(y_true, y_pred)
metrics["precision"] = precision_score(y_true, y_pred, average="weighted", zero_division=0)
metrics["recall"] = recall_score(y_true, y_pred, average="weighted", zero_division=0)
metrics["f1"] = f1_score(y_true, y_pred, average="weighted", zero_division=0)
elif task_type == "regression":
metrics["mse"] = mean_squared_error(y_true, y_pred)
metrics["rmse"] = np.sqrt(metrics["mse"])
metrics["mae"] = np.mean(np.abs(y_true - y_pred))
metrics["r2"] = r2_score(y_true, y_pred)
else:
raise ValueError("task_type must be 'classification' or 'regression'")
metrics["sample_size"] = len(y_true)
return metrics
# Example usage
if __name__ == "__main__":
y_true = pd.Series([0, 1, 1, 0, 1, 0, 1, 1, 0, 0])
y_pred = pd.Series([0, 1, 0, 0, 1, 1, 1, 0, 0, 1])
results = compute_basic_metrics(y_true, y_pred, task_type="classification")
print("Classification Metrics:", results)
Pattern 2: Production-Ready Metrics and KPIs
import logging
import pandas as pd
import numpy as np
from typing import Any, Dict, List
from sklearn.metrics import classification_report, confusion_matrix
logger = logging.getLogger(__name__)
class MetricsAndKPIs:
"""Production-grade metrics and KPIs tracker and calculator."""
def __init__(self, kpi_names: List[str] = None):
self.kpi_names = kpi_names or ["accuracy", "precision", "recall", "f1", "business_roi"]
self.history: List[Dict[str, Any]] = []
logger.info("Initialized MetricsAndKPIs tracker")
def _validate_inputs(self, y_true: pd.Series, y_pred: pd.Series) -> None:
if not isinstance(y_true, pd.Series) or not isinstance(y_pred, pd.Series):
raise TypeError("Inputs must be pandas Series")
if y_true.empty or y_pred.empty:
raise ValueError("Input Series cannot be empty")
if len(y_true) != len(y_pred):
raise ValueError("y_true and y_pred must have matching lengths")
def execute(self, data: pd.DataFrame, target_col: str = "y_true", pred_col: str = "y_pred") -> Dict[str, Any]:
"""Execute metrics calculation on provided DataFrame."""
try:
self._validate_inputs(data[target_col], data[pred_col])
y_true = data[target_col]
y_pred = data[pred_col]
report = classification_report(y_true, y_pred, output_dict=True, zero_division=0)
cm = confusion_matrix(y_true, y_pred).tolist()
f1 = report.get("macro avg", {}).get("f1-score", 0.0)
business_roi = f1 * 1000
result = {
"metrics": {
"accuracy": report["accuracy"]
"precision_macro": report["macro avg"]["precision"]
"recall_macro": report["macro avg"]["recall"]
"f1_macro": report["macro avg"]["f1-score"]
"confusion_matrix": cm
"business_roi": business_roi
}
"metadata": {
"rows_processed": len(data)
"kpi_names": self.kpi_names
"timestamp": pd.Timestamp.now().isoformat()
}
}
self.history.append(result)
logger.info(f"Metrics computed successfully for {len(data)} rows")
return result
except Exception as e:
logger.error(f"Metrics execution failed: {str(e)}")
raise RuntimeError(f"Failed to compute metrics: {e}") from e
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.