Data Science & ML

Select an appropriate model for [problem]:

Problem: [classification, regression, clustering, etc.]
Target variable: [what you're predicting]
Features: [predictors available]
Dataset size: [training samples]
Performance priority: [accuracy, interpretability, speed]
Constraints: [computation, deployment, etc.]

Model selection:
- Problem type classification
- Candidate algorithms (3-5)
- Pros and cons of each
- Complexity vs performance trade-off
- Training time considerations
- Interpretability requirements
- Production feasibility
- Recommended model and rationale
- Alternative models
- Ensemble approach viability
- Validation strategy
- Success metrics

Design feature engineering for [prediction task]:

Task: [what you're predicting]
Raw data: [available data sources]
Domain: [business context]
Model type: [algorithm you'll use]
Current features: [baseline features]
Performance goal: [target metric]

Feature engineering plan:
- Feature creation opportunities
- Transformation suggestions (log, polynomial, etc.)
- Interaction features
- Time-based features (lags, rolling stats)
- Encoding categorical variables
- Text feature extraction (if applicable)
- Dimensionality reduction needs
- Feature scaling approach
- Domain-specific features
- Feature importance expected
- Validation approach
- Feature documentation

Create a hyperparameter tuning plan for [model]:

Model: [algorithm name]
Dataset size: [samples]
Compute budget: [time/resources]
Metric to optimize: [performance measure]
Current baseline: [baseline performance]

Tuning plan:
- Hyperparameters to tune (prioritized)
- Search space for each parameter
- Search strategy (grid, random, bayesian)
- Cross-validation approach
- Evaluation metric
- Early stopping criteria
- Compute time estimate
- Parallelization strategy
- Results tracking
- Overfitting prevention
- Final configuration selection
- Documentation of results

Interpret this machine learning model:

Model: [type]
Target: [what it predicts]
Features: [predictors used]
Performance: [accuracy/metrics]
Stakeholder: [who needs to understand]
Use case: [decision it supports]

Interpretation:
- Global feature importance
- SHAP or LIME analysis approach
- Key driver identification
- Direction of effects
- Interaction effects
- Decision boundary visualization
- Example predictions explained
- Model strengths and weaknesses
- When model works well/poorly
- Business translation
- Actionable insights
- Trust and validation
- Documentation for stakeholders

Design an A/B test for [hypothesis]:

Hypothesis: [what you're testing]
Metric: [primary success metric]
Current baseline: [control performance]
Minimum detectable effect: [smallest worthwhile change]
Traffic available: [sample size potential]
Test duration: [time available]

Test design:
- Null and alternative hypothesis
- Primary and secondary metrics
- Sample size calculation
- Power analysis (typically 80%)
- Significance level (α = 0.05)
- Randomization approach
- Stratification strategy
- Test duration estimate
- Early stopping rules
- Novelty effect considerations
- Analysis plan
- Decision framework
- Implementation checklist
- Rollout plan if successful

Create a validation strategy for [model]:

Model: [type and purpose]
Data: [dataset characteristics]
Deployment: [how model will be used]
Risk: [consequence of errors]
Update frequency: [retraining cadence]

Validation strategy:
- Train/validation/test split
- Cross-validation approach
- Stratification needs
- Temporal validation (if time series)
- Performance metrics suite
- Baseline comparisons
- Error analysis framework
- Bias and fairness checks
- Robustness testing
- Edge case testing
- Model monitoring plan
- Degradation triggers
- Retraining criteria
- Documentation requirements

Compare algorithms for [problem]:

Problem: [description]
Dataset: [size and characteristics]
Metrics: [how you'll evaluate]
Constraints: [requirements]
Baseline: [current approach if any]

Comparison framework:
- Algorithms to compare (4-6)
- Identical preprocessing
- Same train/test split
- Hyperparameter tuning approach
- Performance metrics table
- Training time comparison
- Inference speed comparison
- Model complexity analysis
- Interpretability assessment
- Production readiness
- Strengths/weaknesses
- Recommendation and rationale
- Ensemble potential

Plan the deployment of [ML model] to production:

Model: [type]
Prediction type: [batch, real-time, streaming]
Volume: [predictions per day]
Latency requirement: [response time]
Existing infrastructure: [tech stack]
Team: [engineering resources]

Deployment plan:
- Architecture design
- API specification
- Model serialization format
- Dependency management
- Scaling strategy
- Monitoring and logging
- Performance metrics tracking
- Model versioning
- A/B testing in production
- Rollback procedure
- Retraining pipeline
- Data drift detection
- Alert thresholds
- Documentation
- Timeline and milestones