One 、 What is machine learning (Mechine Learning)？

answer ： Using existing data ( Experience ), To train a model , Use this model to predict the future . Machine learning is the core of artificial intelligence Mechine Learning.
for example ： You and dog egg 7 Gather under the old locust tree , How to make an appointment with Kaihei , The first two dog eggs were 7 spot 10 It's only three minutes . These two times, the dog eggs arrived late 10 Minutes is experience . Then you will judge by your own experience , Will you leave late next time 10 minute , To take advantage of this 10 Minutes to do something meaningful .
For machine learning, it is personification , Let machines have the ability to learn like people .7 The gathering of points is experience for this person , At this point, you need to convert the data into data to the computer ,
People can make judgments through their own way of thinking , Will the dog egg be late next time 10 minute ; For the computer, it is judged by algorithm .
The computer trains a certain model through its own algorithm , Predict the future through the trained model , Judge whether the dog egg will be late next time 10 minute . Whether the here will arrive late 10 Minutes is a binary problem , The machine will calculate a probability , adopt probably To judge .
In machine learning , Data is king . Much more data 、 good , The final prediction will be more accurate .

for example ：y = a*x + b
y1,x1,y2,x2 Can be equivalent to the data
a,b Is the parameter that requires the solution
Prediction is when new x Put it into the formula , Get y Predicted results

The purpose of machine learning is to generate models
The purpose of data mining is to generate indicators

Machine learning falls into two broad categories ：

Ⅰ Supervised machine learning （y It is known. , And right , Have the right result in advance ）

① Return to ：y The range of is positive infinity to negative infinity
The predicted result is a specific value
for example ： Stocks will rise , How much up ？20%、35%

② classification ：y It's a classification number
The result of the prediction is which category
for example ：NBA Whether the Lakers will win the championship in the finals ？ Meeting 、 Can't

Ⅱ Unsupervised machine learning （y It is unknown. , There is no correct result ）

① clustering ： There is no correct answer y Under the circumstances , Sort the data into piles , Birds of a feather flock together .
adopt x1、x2、x3…xn Dimensions , Close 、 Similar ones fall into one category

② Dimension reduction ： When x1、x2、x3…xn, among n When I was very old , In order to solve the equations faster , Reduce x The number of .

Two 、 Corresponding relation

pattern recognition = machine learning
data mining = machine learning + database
Statistical learning ≈ machine learning
Computer vision = The image processing + machine learning
speech recognition = Voice Processing + machine learning
natural language processing = Text processing + machine learning

3、 ... and 、 Common framework

tensorflow and keras similar numpy and scipy The relationship between

Four 、 Supervised machine learning process

Supervised learning ： The data must contain x And the final result y. Divide the data into two parts ,TrainSet and TestSet, among TrainSet Is a data set for training ;TestSet Data set for testing , Machine training phase will not encounter , Also become a new data set .
Supervised machine learning mainly includes classification 、 Regression of these two categories of applications .
Substitute the training set data into the corresponding algorithm ( The formula ) Get the corresponding model ( Parameters ). With the model , Bring test set data into the model , Get the predicted results , And compare and evaluate with the data results in the test set .

5、 ... and 、 Supervised machine learning training process

6、 ... and 、scikit-learn library

SK-Learm