Predictive Analytics – Sparking Water
If you haven’t looked at H2O yet, its probably working spending a bit of time with the software. Apart from the cool stuff you can do with R, there is Spark, and Sparking Water – cool project name. Anyone using Sparking Water with trade data?
CitiBike New York demo and code is available here.
The ML library of algorithms offer the opportunity to use H2O from a number of angles:
- Dynamic Time Warping – does what the client read (e.g. trade ideas, research) impact the trading behaviour? recurring client patterns
- (Hidden) Markov Models – predicting what trade a client might do next
- Collaborative Filtering with ALS – Similar to Netflix, recommend trades, research, trading ideas.
- Logitistic Regression – find the relationship between a binary response variable e.g. if the client hit a price.