問9
頻出度★★★
データマイニングの事例として, 適切なものはどれか。
平成23年度 特別試験 ITパスポート試験 問9
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 9
- ア
正解 The correct answer is ア, “Investigated which products are frequently bought together with a certain product.” Data mining is a technique that applies statistical analysis and machine learning methods to large accumulations of data — such as purchase histories from POS registers or usage histories from membership cards — in order to dig out regularities, correlations, and patterns that people had not noticed in advance (mining = to dig out). The key point is that it does not simply aggregate data to obtain a figure that is already known; it newly discovers hidden relationships. Choice ア is exactly this kind of discovery-oriented analysis: it compares the contents of huge numbers of individual receipts (shopping baskets) and finds association rules about simultaneous purchases, such as the tendency for customers who buy product X to also buy product Y at the same time. This analysis is called basket analysis (association analysis) and is presented as the most representative example of data mining. The rules obtained can be applied directly to store layout and sales promotion: placing X and Y on nearby shelves, offering set sales or cross-purchase coupons, and recommending related products at the register. The decisive difference from mere aggregation processing is that new knowledge about product combinations emerges from the data even without a person specifying a hypothesis, so ア is the appropriate example.
Incorrectイ
Investigating the average monthly sales of a certain product over the past three years is nothing more than simple aggregation (basic statistical processing): the sales amount, an item that is already known, is totaled for each month and divided by the number of items. It can be obtained with a spreadsheet program’s average function or a standard sales report. The form of the required answer is fixed in advance, and no unknown pattern is being discovered within the data, so this cannot be called an example of data mining.
Incorrectウ
Investigating which retail store sold the product with a certain manufacturing serial number is a query operation in which the serial number is used as a key to search the sales history database and retrieve the matching records. It corresponds to a traceability mechanism that traces manufacturing and distribution history backwards, and it is important for recalls and quality control, but it merely locates data that matches a condition and does not discover any new regularity, so it is not data mining.
Incorrectエ
Investigating the product with the highest sales and the product with the highest profit is merely ranking extraction: sales and profit are aggregated per product, sorted in descending order, and the top entries are taken out. This is a basic aggregation-and-sort operation of the kind also performed in ABC analysis, and both the indicator to focus on and the form of the desired answer are specified by a person from the very start. It does not dig out hidden relationships among data, so it does not qualify as an example of data mining.
ここがポイント
- Data mining is a technique that applies statistical analysis and machine learning to large accumulations of data in order to discover regularities, correlations, and patterns that were not anticipated in advance. Simple aggregation (averages, totals), conditional searches, and extracting top entries by sorting are all routine operations whose answer form is decided from the start, so they are not called data mining even when the volume of data handled is large.
- The representative example is basket analysis (association analysis), which finds simultaneous-purchase rules — that product X and product Y tend to be bought together — from receipt-level purchase data, and applies them to shelf allocation, set sales, and recommendations. As related terms, you should also learn the data warehouse, which organizes and accumulates data for analysis; the data mart, which carves out part of it for a particular department; and OLAP, which performs multidimensional aggregation interactively.
用語