A SVM-Based committee machine for prediction of Hong Kong horse racing

Abstract

In Hong Kong and Macao, horse racing is the most famous gambling with a long history. This study proposes a novel approach to predict the horse racing results in Hong Kong. A three-years-long race records dataset obtained from Hong Kong Jockey Club was used for training a support-vector-machine-based committee machine. Bet suggestions could be made to gamblers by studying previous data though machine learning. In experiment, there are 2691 races and 33532 horse records obtained. Experiments focus on accuracy and return rate were conducted separately through constructing a committee machine. Experimental results showed that the accuracy and return rate achieve 70.86% and 800,000% respectively.