Make an improvement in accuracy of validated rows #4
Description
NOTE: after a very brief analysis (TODO: put link to analysis here), the algo in v. 0.0.1 is making a lot of mistakes in identifying valid and invalid taxi trips
A big lift can be made by making a few tweaks to the auditing workflow to appreciate the distribution of continuous values conditioned on the ratecode_id
Some analysis has shown that fitting the data to a linear regression conditioned on ratecode_id teases out some structure of the data and should provide a means of reducing the false positive / false negative rate during validation.
- From the analysis, extract the linear regression coefficients and apply them to the
extreme-numeric?function