--- pipeline_tag: tabular-classification tags: - classification - iris - random-forest - scikit-learn - open-arc metrics: - accuracy - precision - recall - f1 --- # Model Card for InfinitodeLTD/IFSCM-OPEN-ARC Repository: https://github.com/Infinitode/OPEN-ARC/ Project: https://github.com/Infinitode/OPEN-ARC/tree/main/Project-15 ## Model Description The Iris Flower Species Classification Model is a scikit-learn RandomForestClassifier developed as part of Infinitode's OPEN-ARC initiative. It classifies iris flowers as setosa, versicolor, or virginica using sepal and petal measurements. **Architecture**: - **RandomForestClassifier**: `n_estimators=100`, `random_state=42`. - **Framework**: scikit-learn; serialized with joblib. - **Training setup**: Trained on an 80% split of the dataset. - **Model file**: `iris_classifier.joblib`, copied from Project 15 at Git commit `cc4760dbce70898cd05994f6398afb7b3d5f8a0d`. ## Uses - Educational demonstrations of multiclass tabular classification. - Experimentation with the Iris dataset and random forests. - Predicting the three dataset species from measurements in centimetres. ## Limitations - A small, classic dataset does not represent all iris species or real-world conditions. The reported test performance does not guarantee performance on new populations. - Inputs must use the same measurements, units, and column order as the training data. - The project describes its models as learning and research resources, not intended for commercial use. Review the [Project 15 terms](https://github.com/Infinitode/OPEN-ARC/tree/main/Project-15) before use. No Hugging Face license has been declared here because the applicable license for this artifact could not be verified. ## Training Data - Dataset: [Iris Species](https://www.kaggle.com/datasets/uciml/iris). - Content: Sepal and petal lengths and widths, with species labels. - Size: 150 rows, 50 per species, as reported by the notebook. - Preprocessing: Drop missing rows and exclude `Id` and `Species` from features. ## Training Procedure - Train/test split: 80% / 20%, `random_state=42`. - Metrics: Accuracy, and weighted precision, recall and F1. ## Evaluation Results The Project 15 notebook reports the following on its 30-row test split. These are notebook results, not an independent evaluation of the uploaded artifact. | Metric | Value | | ------ | ----- | | Accuracy | 1.0000 | | Weighted precision | 1.0000 | | Weighted recall | 1.0000 | | Weighted F1 | 1.0000 | ## How to Use Install dependencies with `pip install -r requirements.txt`. This example assumes the model is uploaded as `InfinitodeLTD/IFSCM-OPEN-ARC`; change the repository ID if you choose a different name. ```python import joblib import pandas as pd from huggingface_hub import hf_hub_download repo_id = "InfinitodeLTD/IFSCM-OPEN-ARC" # Request the Hub's default download-counting file as well as the model. hf_hub_download(repo_id=repo_id, filename="config.json") model_path = hf_hub_download(repo_id=repo_id, filename="iris_classifier.joblib") model = joblib.load(model_path) flowers = pd.DataFrame( [[5.1, 3.5, 1.4, 0.2], [5.9, 3.0, 4.2, 1.5]], columns=["SepalLengthCm", "SepalWidthCm", "PetalLengthCm", "PetalWidthCm"], ) print(model.predict(flowers)) ``` Only load joblib files from sources you trust: deserialization can execute arbitrary code. The artifact records scikit-learn version 1.9.1; matching the training version is recommended because cross-version loading is unsupported. ## Download Statistics The root `config.json` is provided as the Hub's default download-counting query file. A client must request that file for its request to be counted; downloading only the joblib file will not necessarily increment the counter. GET and HEAD requests can count separately, and cached or offline usage may not generate a request. Counts are not unique users or guaranteed successful model loads. Library-specific Hub counting rules may override the default. See [Hugging Face's documentation](https://huggingface.co/docs/hub/models-download-stats). ## Contact For questions or issues, open a GitHub issue or reach out at https://infinitode.netlify.app/forms/contact.