Machine learning 11 clustering, outlier discrimination

One . clustering

1.1 Estimate the trend of clustering

In fact, clustering with uniform distribution doesn't make much sense
Clusters with random distribution will be particularly numerous , It doesn't make much sense
What is really meaningful is the clustering distribution , Find common classes

1.2 Hopkins Statistics

1.3 Cluster number formulation

Elbow method

1.4 The calculation method of statistics

1.5 Evaluate clustering quality

Bcubed

Profile factor

R Language clustering evaluation

[ Failed to transfer the external chain picture , The origin station may have anti-theft chain mechanism , It is suggested to save the pictures and upload them directly (img-NLlLKsce-1656639913759)(https://upload-images.jianshu.io/upload_images/2638478-db82e380b611c4dc.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)]

1.6 Clustering based on conceptual model

1.6.1 In general EM Algorithm

1.6.2 Examples of concept clusters

1.6.3 Maximum likelihood estimation

Two . Outlier detection

2.1 Methods of detecting outliers

1. Methods based on statistics
2. Proximity based approach
3. Clustering based methods

2.1.1 Statistical methods :

Of univariate outliers Grubb test :

Multivariate outlier test :

2.2 There may be complex situations with multiple clusters

Outliers may form clusters

2.1.2 Histogram method

2.1.3 Judge outliers based on neighborhood

Reference resources :

1. http://www.dataguru.cn/article-4063-1.html