数据挖掘：Data Mining Techniques

该文本介绍了多种机器学习和深度学习的核心概念及其应用。分类（Classification）基于属性预测类别，聚类（Clustering）根据相似性分组，关联规则发现（Association Rule Discovery）基于物品集合的关联性，顺序模式发现（Sequential Pattern Discovery）预测事件的顺序依赖关系，回归（Regression）预测连续值，异常检测（Deviation/Anomaly Detection）识别异常行为。此外，文本还提到了深度学习（Deep Learning）和图学习（Graph Learning）作为机器学习的前沿领域。

文章目录

• 分类分析
• 聚类分析
• 关联规则挖掘
• 序列模式挖掘
• 回归分析
• 异常检测
• 深度学习
• 图学习

Classification 分类

Given a collection of records (training set)

Each record comprises a collection of attributes; among these attributes, one is designated as the class.

Establish a model that class attribute is dependent on the features of other attributes.

The objective is to classify novel records into the most accurate class possible.

该测试集用于评估模型的准确性。通常，给定的数据集会被划分为训练集和测试集，其中训练集用于构建模型，测试集用于验证模型的性能。

Clustering 聚类

Considering a collection of data points, each characterized by a unique set of attributes, and employing a defined measure of similarity among them, the objective is to group data points into clusters such that

• Data points within the same cluster are more similar to each other.

• Data points in distinct clusters are less similar to each other.

• Measures of Similarity:

  • Euclidean Distance when attributes are continuous.
  • Other Problem-specific Measures, including Cosine Similarity, Hamming Distance, and Gaussian Distance among others.

Association Rule Discovery 关联规则发现

• Given a set of records, each containing a certain number of items from a given collection;
• Generate dependency rules that can predict the occurrence of an item based on the occurrences of other items.

Sequential Pattern Discovery 顺序模式发现

• Given a collection of objects, each object is linked to its own timeline of events. Identify rules that elucidate significant sequential dependencies among different events.
• Rules emerge primarily through the identification of patterns. Events within these patterns are subject to timing constraints.

Regression 回归

• Estimate the value of a continuous dependent variable using the values of independent variables, under the assumption of linear or nonlinear dependency relationships. Y = aX + b
• Well-researched in statistics and neural network domains.
• Illustrative examples include:
  • Estimating sales figures for new products based on advertising budgets.
  • Forecasting wind speeds as a function of temperature, humidity, and atmospheric pressure.
  • Time series analysis for stock market index prediction.

偏差与异常检测的识别与处理，该技术在数据处理中具有重要应用价值

• Identify significant anomalies in normal behavior patterns.
  • Applications include:
    • Credit Card Fraud Detection System 信用卡欺诈检测系统
    • Network Intrusion Detection System 网络入侵检测系统

Deep Learning

Graph Learning