Get started

Move from an interaction table to a fair offline comparison. If this is your first time using irspack, follow these guides in order; otherwise, jump directly to the step that matches your current workflow.

Prepare your data

Create stable user and item mappings and choose a split that reflects the product scenario.

Using your own interaction data

Choose a recommender

Establish a cheap baseline, then compare a few models with different assumptions.

Choosing a recommender

Train and evaluate

Work through the first model and evaluate it using ranking metrics.

Train our first movie recommender

Tune a candidate

Apply Optuna-backed tuning after the evaluation pipeline is stable.

Hyperparameter Optimization

• Using your own interaction data
• Choosing a recommender
• Train our first movie recommender
• Evaluation of recommender systems
• Hyperparameter Optimization
• Sampled item evaluation protocol