問題文
A company can train a predictive lead scoring model on three months of data with 45 qualified and 42 disqualified leads, or on two years of data with several hundred of each. Which trade-off should the consultant explain?
選択肢
- Neither period is valid, because the minimum is 40 qualified and 40 disqualified leads per month, and neither of the two data sets reaches that rate consistently.
- The two-year period is invalid because the model scores only the open leads created within the past two years, so training on the same span leaves nothing for the model to be applied to.
- The three-month period is always the better choice, because recent data reflects the current market and the current sales process, and a model trained on stale behavior will mispredict no matter how many records it saw.
- Both meet the minimum, but the more leads included in training, the better the prediction results, so the longer period is likely to perform better unless the sales process changed significantly.