

首页  思维导图  详情



outlier

2016-12-14 10:00:01   0  举报





仅支持查看

AI智能生成

Outlier，也被称为异常值或离群值，是指那些在数据集中偏离大部分观测值的数据点。这些数据点可能由于各种原因而出现，如测量误差、数据录入错误、极端事件等。在统计分析中，处理和识别离群值是非常重要的，因为它们可能会对结果产生显著影响，导致误导性的结论。因此，通常需要采用一些方法来检测和处理离群值，如使用箱线图、Z分数或IQR（四分位距）等。同时，也需要对离群值进行深入的调查和分析，以了解其产生的原因，并决定是否需要删除或替换这些数据点。

模版推荐

作者其他创作

大纲/内容

What

Defination

Why

How

Basic Method

<div style="white-space: pre-wrap; line-height: 1.5; text-align: left; font-size: 14px;" yne-bulb-block="paragraph">Probabilistic and Statistical Models for Outlier Detection</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Extreme Value Analysis</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Univariate</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Probabilistic Tail Inequalities</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Statistical Tail Confidence Tests</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Multivariate</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Depth-base Methods</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Deviation-based Methods</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Angle-based Outlier Detection</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Distance Distribution-based Methods</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Probabilistic Modeling</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Generative model</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Linear Models for Outlier Detection</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Basic linear regression models</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Modeling with Dependent Variables<br></div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Regression Modeling for Mean Square Projection Error</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Principal component method (PCA)</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Proximity-based Outlier Detection</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Cluster-based Methods</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Distance-based Methods</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Nested loop approach</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Cell-based Methods</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Index-based Methods</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Reverse Nearest Neighbor Approach</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Density-based methods</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">LOF: Local Outlier Factor</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">LOCI: Local Correlation Integral</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Histogram-based Techniques</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Kernel Density Estimation</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">High-Dimensional Outlier Detection: The Subspace Method</div>

Unisubspace

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Rarity-based</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Evolutionary algorithms</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Distance-based methods</div>

Multisubspace

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Unbiased</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Random Subspace Sampling</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Aggregation-based</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Selecting High Contrast Subspaces</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Local Selection of Subspace Projections</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Generalized subspaces</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Supervised Outlier Detection</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Fully Supervised</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Rare Class Detection</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Cost Sensitive Learning<br></div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Adaptive Re-sampling</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Semi-Supervised</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Positive and Unlabeled Data</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Difficult Cases and One-Class Learning</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Novel Class Detection</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">One Class Novelty Detection</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Combining Novel Class Detection with Rare Class Detection</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Online Novelty Detection<br></div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Human Supervision</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Active Learning</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Outlier by Example</div>

Specifi Methods

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Outlier Detection in Categorical, Text and Mixed Attribute Data</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Categorical data</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">The generalization of probabilistic models</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">The extension of linear models</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">The extension of proximity-based models</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Binary data</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Subspace Methods</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Novelties in Temporal Transactions</div>

Text data

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Latent Semantic Indexing</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">First Story Detection</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Proximity-based Models</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Probabilistic Models</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Time Series and Multidimensional Streming Outlier Detection</div>

Time Series

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Contextual anomalies: value</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Correlations across time</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Univariate</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Autoregressive Models(RA)</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Moving Average Model (MA)</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Auto-Regressive Moving Average (ARMA)</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Multivariate</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Hidden Variable-based Models</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Correlations across series</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Direct Generalization of Auto-Regressive Models</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Principal Component Analysis  Models</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Supervised Outlier Detection in Time Series</div>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Collective anomalies: unusual shape</div>

Representations

<p style="language:zh-CN;margin-top:4.32pt;margin-bottom:0pt;text-align:left;
direction:ltr;unicode-bidi:embed;vertical-align:baseline;mso-line-break-override:
restrictions;punctuation-wrap:simple"><span style="font-size:12.0pt;font-family:
Calibri;mso-ascii-font-family:Calibri;mso-fareast-font-family:宋体;mso-bidi-font-family:
+mn-cs;mso-ascii-theme-font:minor-latin;mso-fareast-theme-font:minor-fareast;
mso-bidi-theme-font:minor-bidi;color:black;mso-color-index:1;mso-font-kerning:
12.0pt;language:en-US;mso-style-textfill-type:solid;mso-style-textfill-fill-themecolor:
text1;mso-style-textfill-fill-color:black;mso-style-textfill-fill-alpha:100.0%">Numeric Multidimensional Transformation</span></p>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Multidimensional Data Streams</div>

<p style="language:zh-CN;line-height:normal;margin-top:4.32pt;margin-bottom:
0pt;margin-left:0in;margin-right:0in;text-indent:0in;text-align:left;
direction:ltr;unicode-bidi:embed;vertical-align:baseline;mso-line-break-override:
restrictions;word-break:normal;punctuation-wrap:simple"><span style="font-size:
12.0pt;font-family:Calibri;mso-ascii-font-family:Calibri;mso-fareast-font-family:
宋体;mso-bidi-font-family:+mn-cs;mso-ascii-theme-font:minor-latin;mso-fareast-theme-font:
minor-fareast;mso-bidi-theme-font:minor-bidi;color:black;mso-color-index:1;
mso-font-kerning:12.0pt;language:en-US;mso-style-textfill-type:solid;
mso-style-textfill-fill-themecolor:text1;mso-style-textfill-fill-color:black;
mso-style-textfill-fill-alpha:100.0%">Density-based methods</span></p>

<span style="font-size:12.0pt;font-family:Calibri;
mso-ascii-font-family:Calibri;mso-fareast-font-family:宋体;mso-bidi-font-family:
+mn-cs;mso-ascii-theme-font:minor-latin;mso-fareast-theme-font:minor-fareast;
mso-bidi-theme-font:minor-bidi;color:black;mso-color-index:1;mso-font-kerning:
12.0pt;language:en-US;mso-style-textfill-type:solid;mso-style-textfill-fill-themecolor:
text1;mso-style-textfill-fill-color:black;mso-style-textfill-fill-alpha:100.0%">Rare and Novel Class Detection</span>

<div style="white-space: pre-wrap; text-align: left; line-height: 1.5; font-size: 14px;" yne-bulb-block="paragraph">Outlier Detection in Discrete Sequences</div>