codeworks-data/mvp-horse-racing-prediction

Using machine learning to predict HK Horse racing results

Hong Kong Horse Racing Prediction

The aim of this project is to predict the outcome of horse racing using machine learning algorithms.

From RaceBets

Dataset

The dataset comes from Kaggle and covers races in HK from 1997 to 2005.
The data consists of 6,349 races with 4,405 runners.
The 5,878 races ran before January 2005 are used to develop the forecasting models whereas the remaining 471 races ran after January 2005 are preserved to conduct out-of-sample testing.

We have an article explaining our journey through this process. You can find a link below:

• Horse Racing Prediction Part 1
• Horse Racing Prediction Part 2

Documentation

• requirements.txt: list of requirements needed to run this project
• baseline_models.ipynb: notebook containing informations for part 1 on baseline models
• quick_eda_horse_racing.ipynb: notebook with a quick EDA on our dataset
• create_dataset.py and config.py are both used to split our inital data into train and test sets depending on the date of races
• extract_features.py is used to perform feature engineering
• winner/: folder containing all notebooks and ML models to bet on the winner
• placed/: folder containing all notebooks and ML models to bet on placed horses (the Top 3)

winner folder 🏆

Let's have a look about the winner files

• winner_01_lgbm_optim: runs the hyperoptimization for LGBM
• winner_02_train: runs all training processes either for LGBM and deep learning then saves results
• winner_03_show_result: helps us to verify our informations and go deeper about our predictions for a specific month
• winner_04_all_results: consolidates all months with an ensemble model and shows final results
• winner_functions.py: contains the required functions to run those 4 previous notebooks
• model/: contains all saved models from winner_02_train
• result_hyperopt.csv: file with all our optimizations steps

placed folder 🥇🥈🥉

Let's have a look about the placed files

• placed_01_train: runs all training processes for deep learning then saves results
• placed_02_show_result: helps us to verify our informations and go deeper about our predictions for a specific month
• placed_03_consolidated: consolidates all months with an ensemble model and shows final results
• placed_functions.py: contains the required functions to run those 4 previous notebooks
• model/: contains all saved models from placed_01_train and LGBM models from winner_folder

mvp-horse-racing-prediction's Issues

WHERE IS THE STARTER DATA??

WHAT IS THE CREATE DATA FOR...AND HOW TO RUN THE CONFIG AND CREATE DATSET FILES...IN VSCODE..I HAVE PYTHON INSTALLED ADDED TO ENV..AND MY PACKAGES LIKE PANDAS ARE ALSO INSTALLED VIA PIP..CAN SOMEONE HELP PLEASE??

'place' feature

Hi, first and foremost, thank you for the medium article. It has helped me get started with a horse-racing prediction model

mvp-horse-racing-prediction/extract_features.py

Line 90 in 37bc665

df['place'] = list(map(is_place, df['result']))

I have a question regarding your code. From what I understand (on the line above), this 'place' feature is a boolean to track if the horse finished within top 3 in the current race. Wouldn't this potentially be looking forward on the data if included in the dataset?