Table 2.

Summary of notation in user-defined arguments of FIDDLE

Argument	Description
$T$	A positive number specifying the time of prediction; $[0,T]$ is the observation period to consider when processing time-dependent data. The unit of $T$ could be minutes, hours, days, etc., and must be the same as the unit of $dt$.
$dt$	A positive number specifying the temporal granularity (eg, hourly vs daily) at which to resample the time-dependent data. The unit of $dt$ must be the same as the unit of $T$.
${\theta }_1$	A value between 0 and 1 specifying the threshold for the pre-filter step.
${\theta }_2$	A value between 0 and 1 specifying the threshold for the post-filter step.
${\theta }_{\mathrm{freq}}$	A positive number specifying the threshold, in terms of the average number of measurements per time window, at which we deem a variable “frequent” (for which summary statistics will be calculated).
${\{\varphi \}}_{j=1}^K$	A set of $K$ statistics functions (eg, min, max, mean). Each function takes as input 1 or more recordings within a time bin and outputs a single summary statistic. These functions are only applicable to “frequent” variables, as determined by ${\theta }_{\mathrm{freq}}$.
discretize	A Boolean flag (default value: True) specifying whether features with numerical values are kept as raw values or discretized into binary features.
discretization_encoding	A string specifying how numerical values are encoded into binary features after discretization. Possible values are: “one-hot” (default) and “ordinal.” This argument is ignored and should not be used when discretize=False.

Note: FIDDLE: Flexible Data-Driven Pipeline.