How constancedongg/horse-racing works
Main Components
• Data Sources:
• The raw input files: training.csv, testing.csv, race-result-horse.csv, and race-result-race.csv (these represent the data foundation from Kaggle and the Hong Kong Jockey Club).
• Data Processing & Preprocessing:
• Preprocessing steps that the Notebook carries out (including cleaning, feature engineering, and merging of race and horse data).
• Machine Learning Models:
• Classification algorithms: Logistic Regression, Naïve Bayes, SVM Classifier, Random Forest.
• Regression methods: SVR and Gradient Boosting Regression Tree (GBRT model).
• Prediction Outputs:
• The prediction CSV files, e.g., lr_predictions.csv, nb_predictions.csv, svm_predictions.csv, rf_predictions.csv, reg_prediction.csv, and nb_predictions.csv.
• Analysis & Betting Strategy:
• The Notebook also contains evaluation, model performance comparisons, and concluding recommendations about betting strategies.
Relationships and Interactions
• Data Flow:
• From Data Processing, output flows into each ML Model component.
• Evaluation & Strategy:
• A feedback loop from Prediction Outputs (and optionally the processed data) to an Analysis/Strategy component where results are compared and insights are drawn.
• Notebook as an Orchestrator:
• Jupyter Notebook (Horse_Racing.ipynb) as the central orchestration hub that ties data ingestion, preprocessing, model building, prediction generation, and strategy development together.
Architectural Patterns
• Data Pipeline Pattern: Separate inputs (raw data) from processing steps (preprocessing, feature engineering, model training/evaluation).
• Modular ML Pipeline: Each algorithm (classification and regression) can be depicted as a modular “component” with specific inputs and outputs.
Relevant Technologies and Tools
• Jupyter Notebook (Python environment) as the main development and narrative tool.
• Machine Learning libraries (implied: scikit-learn or similar, based on the algorithms mentioned).
• CSV files denote direct file-based persistence for both input data and predictions.
Additional Detail
• Although the project is not a full-stack application, it benefits from a clear separation into sub-systems (data ingestion, processing, modeling, predictions, analysis) which helps in both understanding and potentially extending the system.