Machine Learning (ML) Dataset

A Machine Learning (ML) Dataset is a dataset used in a machine learning task.

• Context:
  • Task Input: Raw Data Records, Data Schema
  • Task Output: Processed Dataset, Feature Vectors
  • Task Performance Measure: Dataset Quality Metrics such as:
    • Data Completeness Score
    • Feature Quality Score
    • Distribution Balance Score
  • ...
  • It can (typically) contain learning records for model training and evaluation.
  • It can (typically) serve as a task input to a machine learning algorithm.
  • It can (typically) be processed by a data preparation task before use.
  • ...
  • It can range from being a Numerical ML Dataset to being a Categorical ML Dataset, depending on its data type.
  • It can range from being an Unlabeled ML Dataset to being a Labeled ML Dataset, depending on its label availability.
  • It can range from being a Single-Predictor Learning Dataset to being a Multi-Predictor Learning Dataset, depending on its feature dimensionality.
  • It can range from being a Training Dataset to being an Evaluation Dataset, depending on its usage purpose.
  • It can range from being a Static ML Dataset to being a Streaming ML Dataset, depending on its data collection method.
  • It can range from being a Balanced ML Dataset to being an Imbalanced ML Dataset, depending on its class distribution.
  • It can range from being a Small Scale Dataset to being a Large Scale Dataset, depending on its data volume.
  • It can range from being a Clean ML Dataset to being a Noisy ML Dataset, depending on its data quality.
  • ...
• Examples:
  • Standard ML Datasets, such as:
    • Classification Datasets, such as:
      • MNIST Dataset for digit recognition.
      • CIFAR Dataset for image classification.
      • Reuters Dataset for text classification.
    • Regression Datasets, such as:
      • Boston Housing Dataset for price prediction.
      • California Housing Dataset for real estate analysis.
      • Weather Dataset for temperature forecasting.
  • Special Purpose Datasets, such as:
    • Time Series Datasets, such as:
      • Stock Market Dataset for financial prediction.
      • Sensor Data Collection for anomaly detection.
    • Graph Datasets, such as:
      • Social Network Dataset for community detection.
      • Chemical Structure Dataset for molecular property prediction.
  • Domain-Specific Datasets, such as:
    • Healthcare Datasets for medical diagnosis.
    • Genomic Datasets for gene expression analysis.
    • Industrial Datasets for process optimization.
  • ...
• Counter-Examples:
  • User Profile Dataset, which is for user management rather than learning.
  • System Log Dataset, which is for monitoring rather than learning.
  • Reference Dataset, which is for lookup rather than learning.
  • Backup Dataset, which is for data preservation rather than learning.
• See: Data Record Set, User-Interaction Data, Feature Engineering, Data Preprocessing, Model Training, Machine Learning System, Dataset Quality Assessment.

---

---

References

2017

• (Sammut & Webb, 2017) ⇒ (2017) "Data Set". In: Sammut & Webb, 2017.
  • QUOTE: A data set is a collection of data used for some specific machine learning purpose. A training set is a data set that is used as input to a learning system, which analyzes it to learn a model. A test set or evaluation set is a data set containing data that are used to evaluate the model learned by a learning system. A training set may be divided further into a growing set and a pruning set. Where the training set and the test set contain disjoint sets of data, the test set is known as a holdout set.

---