ML Feature Engineering
Derive new meaningful features from raw data for machine learning models
Prompt template
You are an experienced machine learning engineer. Help me with feature engineering for the following ML project. Problem statement and goal: [PROBLEM – e.g., predicting customer churn within the next 30 days] Available raw data: [RAW_DATA – description of existing columns, e.g., transaction history, login timestamps, support tickets, demographic data] Target algorithms: [ALGORITHMS – e.g., Gradient Boosting, Logistic Regression, Random Forest, neural networks] Programming environment: [ENVIRONMENT – e.g., Python with scikit-learn and pandas, PySpark for big data] Please provide: 1. At least 15 concrete feature ideas with justification for their relevance to the target problem 2. Complete code for computing each feature 3. Recommendations for feature scaling and encoding (One-Hot, Target Encoding, normalization) 4. Strategies for handling missing values and outliers in the feature engineering context 5. Methods for feature selection and importance ranking (e.g., SHAP, correlation analysis)
How to use this template
Copy the prompt, replace the placeholders in square brackets with your details and paste it into the AI assistant of your choice. The more precise your details, the better the result.
Tips for better results
- Describe the target variable precisely, as this determines which features might be relevant.
- Mention whether temporal order matters — time-based features require special techniques like lag features.
- Indicate whether model interpretability is important, as this influences the choice of feature engineering methods.
Suitable open models
The selection names current open models as a starting point for your own tests. It is not a measurement for this template.
Keywords: #feature-engineering #machine-learning #scikit-learn #modeling