Model-performance is a report that shows a performance of a predictive model in different business segments. With the help of this report the customer can see bias in different segments that can be examined by Data Analysts. Those can lead to fine tuning and adaptation of the model that can be tracked over time.
Key components in the report
- Model filter - choose the model to show
- Segment type filter - choose the segment type to review
- Segment filter - choose the segment
- Actual vs. predicted radio button - choose if its predicted report or actual report from conversion data
- Training Records/Scored Records - number of records represented in the report
- Time Period - time period of the report
- Average Conversion - average of conversion of incidents
- Total Converted - number of converted incidents
- Total Revenue - total revenue summarized
- Converted vs. Incident per Score graph - bar charts that shows the distribution of converted rows vs. incident rows of data
- Score Range row - shows the predictive scores ranges per report buckets
- % of Converted row - percent of incidents that converted
- % of Revenue row - percent of revenue from the converted rows
- % of Records row - percent of incident rows that are categorized per each bucket
- Conversion Lift row - lift of conversion in the bucket
How to create a report
Currently, these report can only be produced offline by our data analyst team. Users can reach out to their CSM to request this, and upon creation it will show in their Leadspace On-Demand account.