Loud ML provides a full Feature DSL (Domain Specific Language) based on JSON to define how data must be organized in memory in order to produce the right model and predictive output with your data set.
Think of the Feature DSL as an abstract syntax consisting of two types of structures:
- Timeseries types
Timeseries types aggregate data based on a particular field, or script,
numericfeatures. These features can be used by themselves to predict their trend in the near future.
- Organized types
- Organized features wrap other compound features and are used to combine multiple categorical and numerical properties in a logical fashion. These features are used to discover patterns in the data.