一、为什么要挖掘数据
1.数据丰富但信息贫乏
2.电脑便宜且强大
3.大量数据超出人的理解范围
4.数据收集与储存的速度快（science）
5.传统工具对原始数据不可行（science）
6.数据仓库有大量数据（industy）
7.企业竞争太大（industy）
注1：Data mining turns a large collection of data into knowledge
二、什么是数据挖掘（DM）
Data Mining: process of semi-automatically analyzing large databases to find patterns that are:
Valid: hold on new data with some certainty
Novel: non-obvious to the system
Useful: should be possible to act on the item
Understandable: humans should be able to interpret the pattern
注1：Data Mining is the progress of discovering interesting patterns from massive amounts of data（数据挖掘是从大量数据中发现有趣的模式的过程）
三、:什么样的数据可以被挖掘What Kinds of Data Can be Mined?
1. Database Data
2. Data Warehouses
3. Transaction Data（事物数据）
4. Spatial-Temporal Data（时空数据）
5. Graph and networked data
6. Hypertext and multimedia data：Text, image, video, and audio data
7. Time-related sequence data：Historical records Stock exchange eg.
8. Data Stream:Video surveillance
四、什么样的模式可以被挖掘What Kinds of Patterns Can be Mined?
1. Class/Concept Description(类/概念描述): Characterization and Discrimination
Data Characterization（数据特征化）:
Tools: Statistical measures and plots
Outputs: Pie charts, bar charts, curves, multi-dimensional data cubes, and multi- dimensional tables, generalized relations.
Data Discrimination（数据识别/区分）：
Comparison of the general features of the target class data objects against the general features of objects from one or multiple contrasting classes.（将目标类数据对象的一般特性与来自一个或多个对比类的对象的一般特性进行比较）
Outputs: comparative measures that help to distinguish between the target and contrasting classes（有助于区分目标类和对比类的比较度量）
2. Mining Frequent Patterns, Associations and Correlations
Mining frequent patterns leads to the discovery of interesting associations and correlations within data（挖掘频繁模式可以发现数据中有趣的关联和关联）
3. Classification（分类） and Regression（回归） for Predictive Analysis
Classification
Training data: data objects with class labels are known（具有类标签的数据对象是已知的）
Output: Find a model that describes and distinguishes data classes or concepts
Regression Analysis
Regression models continuous-valued functions
4. Cluster Analysis（聚类分析）
Points (objects) that are “close” in the attribute (feature) space are assigned to the same cluster（属性(特性)空间中“接近”的点(对象)被分配给同一个集群）
5. Outlier Analysis（异常分析）
In some applications such as fraud detection, the rare events can be more interesting than the more regularly occurring ones. The analysis of outlier data is referred to as outlier analysis or anomaly mining（在一些应用程序中，例如欺诈检测，罕见事件可能比经常发生的事件更有趣。异常数据的分析称为异常分析或异常挖掘）
Outliners may be detected using statistical tests, or using distance measures（可以使用统计测试或使用距离度量来检测外线程）
注1：A big data-mining risk is that you will “discover” patterns that are meaningless（一个大的数据挖掘风险是您将“发现”毫无意义的模式）
注2：A pattern is interesting if it is valid on test data with some degree of certainty, novel, potentially useful（如果模式在测试数据上是有效的，并且具有一定的确定性、新奇性和潜在的有用性，那么模式就是有趣的）
五、使用了哪些技术?Which technologies are used?
machine learning, statistics, artificial intelligence, databases, visualization（可视化） but more stress on：
Scalability of number of features and instances 特性和实例数量的可扩展性
Stress on algorithms and architectures whereas foundations of methods and formulations provided by statistics and machine learning 强调算法和架构，而统计和机器学习提供的方法和公式的基础
Automation for handling large, heterogeneous data 用于处理大型异构数据的自动化
注1：Data mining, as a highly application-driven domain, has incorporated knowledge from many other domains（数据挖掘作为一个高度应用驱动的领域，吸收了许多其他领域的知识）
六：什么样的应用是目标?What kinds of applications are targeted?
1.Web Mining:
Decide the importance of pages: PageRank algorithm
2.Market and Sales
3.Medicine:
Disease outcome, effectiveness of treatments
4.Molecular/Pharmaceutical:
Identify new drugs
5.Scientific data analysis:
Identify new galaxies by searching for sub clusters
注1：Data mining has many successful applications, such as business intelligence, Web search, bioinformatics, health informatics, finance, digital libraries, and digital governments（数据挖掘有许多成功的应用，如商业智能、网络搜索、生物信息学、卫生信息学、金融、数字图书馆和数字政府）
网友评论