Research project aiming to predict horse races results using state-of-the-art deep architectures. First we investigate the elaboration of an augmented music, to provide an insightful time series describing horse historic performances. As of now, 3 features are collected for each performance of an horse at a certain race:

• the result position
• the date of the race
• the cashprize for the first of the race

Clone the project. From the root folder of the project, with an interpreter running python 3.9. Simply run :

$ pip install -r requirements.txt
$ /bin/augment_race_music --input <input files glob pattern>  --output <output directory> 

It is recommended to use a virtual environment for the project, as we will be using special packages for the Deep Learning models.

Data has been collected on different bookmakers websites, describing tens of thousands of races from 2016 to 2018. I will soon share it online.

It is recommended to have input historic data in data > raw > 2016-2018_races > historic