dataScore: 88/100For ChatGPT

Complete Predictive Modeling Strategy (ML)

Establishes an end-to-end machine learning strategy with algorithm selection and evaluation metrics

Fill in the variables

Your personalized prompt

You are an expert data scientist in machine learning. Design a complete predictive modeling strategy.

# PROJECT CONTEXT
Dataset description: {{dataset_description}}
Available predictive variables: {{variables_predictives}}
Target variable to predict: {{variable_cible}}
Business constraints: {{contraintes_metier}}

# EXPECTED STRATEGY

## 1. Problem Definition
- Problem type (classification, regression, clustering, etc.)
- Business success metrics
- Performance baseline to beat
- Use case and expected value

## 2. Data Preparation
- Cleaning and imputation strategy
- Recommended feature engineering
- Categorical variable encoding
- Normalization/standardization if needed
- Imbalance handling (if classification)
- Split strategy (train/validation/test)

## 3. Algorithm Selection
Recommend 3-5 candidate algorithms:
- Algorithm name
- Choice justification
- Advantages for this use case
- Potential drawbacks
- Key hyperparameters to tune

## 4. Evaluation Metrics
For {{variable_cible}}, define:
- Primary metric (e.g., F1-score, RMSE, AUC-ROC)
- Secondary metrics
- Acceptable performance thresholds
- Validation method (cross-validation, etc.)

## 5. Experimentation Plan
```
Phase 1: Simple baseline
- Algorithm
- Configuration
- Expected result

Phase 2: Intermediate models
- [Model list]

Phase 3: Optimization
- Hyperparameter tuning (Grid Search, Bayesian)
- Feature selection
- Ensemble methods if relevant
```

## 6. Overfitting Management
- Regularization techniques
- Cross-validation
- Early stopping strategy
- Learning curves analysis

## 7. Interpretability
- SHAP values or LIME for explanations
- Feature importance
- Error analysis
- Key visualizations

## 8. Production Deployment Plan
- Model saving format
- Monitoring strategy (drift detection)
- Re-training frequency
- Version management
- Prediction API (architecture)

## 9. Python Starter Code
Provide a structured notebook with:
```python
# Imports
# Data loading
# Quick EDA
# Preprocessing pipeline
# Baseline training
# Evaluation
# Model saving
```

Adapt your recommendations to constraints: {{contraintes_metier}}

Go further: optimize this prompt →