YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
SteamPulse 🎮
Machine Learning for Steam Game Analysis
Overview
SteamPulse is a machine learning project that explores patterns in the Steam game catalogue using game metadata, text features, and structured information.
This repository hosts the trained Logistic Regression release candidate developed for the SteamPulse project.
Model Details
- Task: Binary classification
- Target:
hit - Algorithm: Logistic Regression
- Regularization: L2
- Class weighting: Balanced
- Maximum iterations: 5,000
- Combined feature count: 45,518
- Decision threshold: 0.0191066
The model uses text representations of game titles and descriptions, multilabel genre and category features, and structured game metadata.
The target represents a proxy for cumulative observed Steam recommendation engagement. It does not directly measure revenue, profitability, or verified commercial success.
Validation Results
The following metrics are recorded in the saved model artifact.
| Metric | Score |
|---|---|
| ROC-AUC | 0.8451 |
| PR-AUC | 0.4467 |
| Precision | 0.3503 |
| Recall | 0.6024 |
| F1 score | 0.4430 |
The model uses a decision-function threshold of approximately 0.01911. This threshold is not a probability cutoff.
Intended Use
SteamPulse is intended for experimentation, exploratory analysis, and investigation of patterns in Steam game metadata.
Predictions should be interpreted as model estimates of the defined engagement-related target, not guarantees of future game performance or commercial success.
Limitations
- The target is a proxy for cumulative observed recommendation engagement, not verified commercial success.
- Current metadata may differ from the metadata available when a game was originally released.
- Model performance depends on the training data and evaluation methodology.
- Predictions may be less reliable for games that differ substantially from the training data.
Technical Information
The saved artifact contains the trained estimator, preprocessing objects, feature schema, target definition, decision threshold, validation metrics, and software-version information.
The artifact was created using Python 3.13.9, scikit-learn 1.7.2, NumPy 2.3.5, pandas 2.3.3, and SciPy 1.16.3.