[advanced data mining technology] Introduction to advanced data mining technology

Functional classification of data mining technology

Descriptive data mining （Descriptive Data Mining（Unsupervised Learning Unsupervised learning , No target value is required ））

Association Rules（ Association rules ）

Find out which events often occur together ,

Example ： Amazon Amazon.com

People who buy these two books will generally buy the following books , According to past purchase records , Find the connection . Most e-commerce companies have recommendations , The technology behind it is related technology .

This technology is to find those events that happen at the same time .（ Buy two books at the same time ）

Generally used in e-commerce , Virtual store , You can make recommendations directly in the purchase process （ Just like the battery and flashlight in front of each other ）

Algorithm ：

Apriori

FP-Growth

Sequential Patterns( Sequence pattern )

Find out which events often appear in sequence

This technology is to find out what products will be purchased again after purchasing a certain product . There is a time order relationship

Generally used in physical stores , Like the bookstore above , After paying for the book , You can predict the goods you may buy next time , Print out this type of discount coupon , Attract customers to buy again , Or consumption . Because the sales are not at the same time ,up_selling. There are recommendations in chronological order

Algorithm ：

ApriorAll

Cluster Analysis（ Clustering analysis ）

Find out the internal structure between data

Most of the fields of customers of the same type will be the same , Fields of different classes differ greatly

The above figure is a bank that hopes to cluster the investment propensity of consumers by income and age .

One point represents one customer , If the points are dense and concentrated, a group will be formed . Can we see that there is 3 Species group . We can make some corresponding marketing strategies for different customer groups

The first group has low income , Middle and upper age （ Investigation found , It may be workers , High risk , Insurance Strategy , The insurance cost is low .）

The second group has a high income , Older （ Old and golden , Prefer capital guaranteed investment ）

The third group has a medium to high income , Younger （ Rich second generation , Tend to high-yield and high-risk investment , The risk tolerance is relatively large ）

Algorithm ：

Hierarchical clustering ：Single Linkage,Aberage Linkage,Complete linkage

Separate clusters ：K-means

Kohonen self Organizing Maps（SOM）

Two-Step First determine the number of clusters , How to get a group （ You can get several groups ）

Predictive data mining （Predictive Data Mining（Supervised Learning Supervised learning ））

Classification（ classification ）

The category to which the forecast data belongs

You need to input variables , Target variables are also required .

Predict whether customers are interested in car models

Customer number Input properties Result properties

Input Attributes Input factors ：Car（ The type of car you bought ） Age（ Age ） Children（ There are several children ）

New customers ：car=sedan Age=35 To make predictions See if you have any preferences

The result variable must be a categorical variable

Algorithm ：

Bayes Net

Logistic Regreession （ Logical regression ）

Decision Tree （ Decision tree ）

Neural Network（ neural network ）

Support Vector Machine（ Support vector machine ）

Knearest Neighborhood（K- Nearest neighbor ）

Prediction（ forecast ）

The value corresponding to the predicted data

There should be input variables （ reason ）, The result field must be numeric

For example, predict the annual revenue of customers ,

Location, The location of the house Type Type of house Miles And school location SF House size CM How many houses in the community

forecast

Input field Location=Rural Miles=3 SF=1500

Result fields ：Home Price

Linear Regression（ Linear regression ）

Time Series（ The time series ）

Decision Tree（ Decision tree ）*

Neural Network（ neural network ）*

Support Vector Machine（ Support vector machine ）*

K-Nearest Neighborhood（K- Nearest neighbor ）*

1. 1. 1. Introduction to data mining related websites -KDnugets&Kaggle

KDnugets： Can provide a lot of things , There are data sets , The latest information will be sent to you after registration

Data UCI The data is easy to use . Data sets from all walks of life ,

Kaggle： A bridge between enterprises and data scientists , Enterprises can send data to scientists for analysis .

There are enterprise requirements

Enterprises hand over demand , Compete in the form of competition . It can provide a large amount of data at the enterprise level . It's very large Valued Shoppers（ value ）, More authentic data .

Data Castle, kordsa Kesci（ There is training ）（ China's Kaggle） It also provides data

Positioning of data mining 、 expectation 、 With the establishment of data mining team

Machines replace labor , The first industrial revolution

Data mining middle managers , Will be replaced by Intelligent Automation

Prospects of data mining ：

Time magazine lists data mining as 21 One of the five emerging industries in the 21st century , Data mining is of great importance in business

The future marketing focus will shift from products to customers

1. Customers may be robbed by better services from competitors
2. Whoever has the most knowledge about customers has the most capital
3. Know more about customers , The more we can deepen the uniqueness of the brand , The competition is stronger

Only by converting data into knowledge , Knowledge becomes action , To turn action into profit

Data mining and fortune telling

Data Mining data mining （ Modern fortune telling ）

Data：Attributes

Algorithm：Classification,Clustering,Association,…（ Various algorithms ）

Predict Future Trends（ Future trends ）

Fortune-telling Fortune-telling

Data： The eight words of birth , Face , Palm ,…

Algorithm： Ziwei's number , Four column tweet …

Predict Future fortune

Data mining can predict the future , Reduce fear .

How to carry out data mining ：

Short term data mining requires tools , Long term data mining requires self-development

The team's best case group needs 3 people of the same race

One kind of person is Project Manager（ managers ）

One kind of person is CRM（Data Mining） People who （ data mining ）

One kind of person is IT（Database） People who （ The person who provides the data ）

Data analysis certification process ：

Data Miner1. Given field Attributes And data Data, It works DATA Mining Tools Tools to analyze , Get the results Mining Results

Data Analyzer2. Given field Attributes（ Influencing factors ）, Can operate database Databases,DATA Mining Tools Tool analysis can get and interpret results Mining Results

Business Analyzer3. Given the subject , Can operate database Databases, Tools to analyze DATA Mining Tools Get and interpret the results Mining Results