We have a system for betting on chess matches. Customers bet on specific matches and we gather information about their betting history. The goal is to make the best recommendations for the customers. I will try two approaches:
- neural network,
- matrix factorization, and compare them.
On Windows machines set HADOOP_HOME env variable to directory with winutils binary
- Matrix factorization
- after new data comes in algorithm needs to recount values
- Nearest neighbours
- Deep learning: https://spark.apache.org/docs/latest/ml-classification-regression.html#multilayer-perceptron-classifier
- needs to be calculated for each user. Some generalzation required?