問65
頻出度★★☆
AIにおける機械学習の学習方法に関する次の記述中のa～cに入れる字句の適切な組合せはどれか。 教師あり学習は，正解を付けた学習データを入力することによって，[ a ]と呼ばれる手法で未知のデータを複数のクラスに分けたり，[ b ]と呼ばれる手法でデータの関係性を見つけたりすることができるようになる学習方法である。教師なし学習は，正解を付けない学習データを入力することによって，[ c ]と呼ばれる手法などで次第にデータを正しくグループ分けできるようになる学習方法である。
令和6年度 ITパスポート試験 問65
テクノロジ系基礎理論 › 情報に関する理論
This question's explanation and your answers so far.
解説
問 65
Incorrectア
It sets a: regression and b: classification, swapping a and b. [ a ] should be “dividing into multiple classes” = classification, not regression. Regression is a method that finds relationships (continuous values), so it should come at [ b ]; the correspondence is reversed, so it is wrong.
Incorrectイ
It sets a: clustering and c: regression. Clustering is an unsupervised-learning method that does not use correct-answer labels, so it is unsuitable for [ a ] in supervised learning. Also, [ c ] should be a grouping method = clustering; regression cannot explain “grouping”, so it is wrong.
- ウ
正解 The correct answer is “ウ” (a: classification, b: regression, c: clustering). Supervised learning uses training data labeled with correct answers. [ a ] is a method that “divides unknown data into multiple classes”, so “classification”, which assigns data into pre-prepared classes (categories), fits. [ b ] is a method that “finds relationships among data”; “regression”, which finds the continuous correspondence (numerical trend) between input and output, fits. Unsupervised learning gives no correct-answer labels, and [ c ] is a method that “gradually becomes able to correctly group the data”, so “clustering”, which automatically forms groups (clusters) from the similarity among data, fits. Therefore “ウ”, the combination a = classification, b = regression, c = clustering, is correct.
Incorrectエ
a: classification is correct, but b: clustering and c: regression are wrong. [ b ] should be “finding relationships among data” = regression, not clustering (unsupervised grouping). [ c ] should be “grouping” = clustering; regression does not fit.
ここがポイント
- The representative methods of supervised learning are “classification” (assigning data into discrete classes) and “regression” (predicting continuous values and relationships); the representative method of unsupervised learning is “clustering” (automatically grouping by similarity without labels).
- Distinguish by matching keywords to methods: “dividing into multiple classes” → classification, “finding relationships (continuous values)” → regression, “gradually grouping without correct answers” → clustering.
用語