引導式測驗 · 中等 等級
市場籃分析和經常一起購買的測驗
了解支持度、置信度和提升如何描述共同購買,方向如何改變置信度,以及為何關聯規則無法建立因果關係。
問題 1 的 8
What does support for the rule X to Y measure?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- What does support for the rule X to Y measure?
- The count of baskets with both X and Y divided by all baskets
- The count of baskets with X divided by all baskets
- The count of baskets with both items divided by baskets containing X
- The change in Y purchases caused by introducing X
- How is confidence for X to Y calculated?
- Joint count divided by all transactions
- Count of baskets with X divided by count of baskets with Y
- Joint count divided by the count of baskets containing X
- Confidence of Y to X multiplied by lift
- What does lift above one indicate?
- The joint count exceeds the minimum support threshold
- Y occurs with X more often than its overall baseline rate
- Every basket containing X also contains Y
- Y has no purchases outside baskets containing X
- Why can reversing X to Y change confidence?
- The confidence denominator remains all transactions regardless of direction
- Reversing the rule changes the pair’s joint support
- The overall transaction population changes
- The denominator changes to baskets containing the new antecedent
- What stays the same when a two-item rule is reversed?
- Joint support and pairwise lift
- Confidence and the frequency of the antecedent
- The antecedent frequency used as confidence denominator
- The confidence value in both directions
- Why can high confidence be misleading?
- A high confidence value means X is rare
- Confidence already adjusts for the baseline rate of Y
- The consequent may already be very common overall
- A high value proves that X caused Y
- What does a market-basket association establish?
- A causal effect of displaying one item beside the other
- A co-occurrence pattern in the analyzed transactions
- A preference shared by every customer in the sample
- A pattern that must persist across future periods and stores
- How can a retailer test whether a widget caused a change?
- Compare confidence before and after changing the widget
- Treat a large lift as a causal estimate of the widget
- Compare only the customers who clicked the recommendation
- Run a controlled experiment and compare outcomes