Machine learning 10 belief Bayesian classifier

One . Bayesian overview of belief

Two . clustering

You can mark every variable , How many different values , How many dimensions are there , In space, if the distance is closer , Then they can be regarded as the same kind , It can be marked by clustering algorithm .

2.1 Application scenario of clustering : Looking for quality customers

2.2 Clustering application scenarios : Recommendation system

2.3 Clustering application scenarios : Community discovery

2.4 Clustering application scenarios : The tree of biological evolution

2.5 Clustering application scenarios : The special meaning of outliers

2.6 R Of dist function

2.7 Data centralization and standardization transformation

3、 ... and . Hierarchical clustering

3.1 The method of calculating the distance between classes

1. The shortest distance method
2. The longest distance method
3. Middle distance method
4. Class average method
5. Barycenter method
6. The sum of squares of deviations

R Medium hclust() function :

3.2 Determination of the number of classes

Four . Dynamic clustering : K-means Method

K-means Easier to move Hadoop Of MR On

4.1 R Medium kmeans() function

4.2 K-means Advantages and disadvantages of the algorithm

K The mean value is better for spherical clusters , If it is the plane shown in the following figure , The effect is not ideal

5、 ... and . Technology based on representative points ： K Central clustering

K Implementation of the central method :PAM

6、 ... and . Density based approach : DBSCAN

Some concepts

7、 ... and . CLARA（ Big data processing ）

Reference resources :

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