We disabled the sales chasing detection in the reporting.ratio_stats table when it was converted to a Python model in #422. The reasons to disable it were enumerated in that PR, but to recap:
The detect_chasing() function specifically did not play well with the Spark environment for some reason (probably using ECDF()).
The results of the chasing detection were not heavily used anywhere.
The detect_chasing() function combines two methods using an AND operator. At least one of those methods was highly sensitive to/created lots of false positives for areas with extremely homogenous housing stock.
However, the chasing detection output is valuable. We should update the ratio_stats Python model to reimplement chasing with improved methods. "Improved" here could mean many things, e.g.:
Simply changing some of the parameters set by default for the original methods.
Trying to refine or tweak the existing methods
Creating entirely new methods, incorporating them into assesspy, and then integrating them into the Python model
We disabled the sales chasing detection in the
reporting.ratio_statstable when it was converted to a Python model in #422. The reasons to disable it were enumerated in that PR, but to recap:detect_chasing()function specifically did not play well with the Spark environment for some reason (probably usingECDF()).detect_chasing()function combines two methods using an AND operator. At least one of those methods was highly sensitive to/created lots of false positives for areas with extremely homogenous housing stock.However, the chasing detection output is valuable. We should update the
ratio_statsPython model to reimplement chasing with improved methods. "Improved" here could mean many things, e.g.:assesspy, and then integrating them into the Python model