問4
頻出度★★★
データマイニングとは，データベースに蓄積されている大量の生データに対し，統計やパターン認識などの手法を用いることによって，認識されていなかった規則性や関係性を導き出す技術である。データマイニングの応用分野として，最も適切なものはどれか。
平成21年度 秋期 ITパスポート試験 問4
ストラテジ系企業活動 › 業務分析・データ利活用
This question's explanation and your answers so far.
解説
問 4
- ア
正解 The correct answer is ア, “Recommending products tailored to each customer.” As the question itself defines, data mining is a technique that uses statistics and pattern recognition to dig out “regularities or relationships that have not been recognized before” from large volumes of raw data. By analyzing POS data and purchase histories, you obtain association rules such as “people who bought product A also tend to buy product B” (association analysis, also known as basket analysis) and customer clustering results, so you can recommend to each customer the product they are likely to buy next. Such recommendation rules were not known to anyone in advance; they are knowledge discovered from the data for the first time, which is exactly a typical application area of data mining (CRM and recommendation).
Incorrectイ
“Allocating manufacturing machines based on a production plan” means finding an optimal allocation from known conditions: an already fixed production plan plus the capacity and constraints of the machines. It deals with optimization and scheduling (production management, the field of operations research), not with discovering unknown regularities from accumulated data.
Incorrectウ
“Aggregating sales by store and by product category” is a simple aggregation that totals figures along axes decided in advance (store, product category). It belongs to the routine aggregation and reporting of OLAP and data warehouses, and the result is merely numbers along known viewpoints — how much each category made at each store — so it does not discover any new regularity or relationship.
Incorrectエ
“Extracting premium customers based on cumulative purchase amount” only sorts and filters by a criterion a person set in advance (a customer is premium if the cumulative amount reaches some threshold). It is a search-and-extract operation that pulls out records matching a condition, not a process that derives unknown patterns from the data itself. Note that an analysis that newly discovers which purchasing behaviors make a customer likely to become a premium customer would be data mining, but this choice stops at extraction by a known criterion.
ここがポイント
- Data mining = a technique that uses statistical methods and pattern recognition to discover regularities, correlations, and patterns in large accumulated data that people have not yet noticed. Typical examples are recommendation based on association analysis (basket analysis), customer clustering, and churn prediction.
- “Totaling along known axes (aggregation, OLAP),” “filtering by known conditions (search and extraction),” and “optimally allocating under known constraints (scheduling)” are all not data mining. The decisive test is whether new knowledge is discovered.
用語