简单英语+机器学习的主要过程

 Hey guys, before we are talking the algorithm and process want to emphasize the common sense in the machine learning world.
 This is the data decide the limitation for the performance of the model. And more different algorithm is just want to touch the Top of the performance or the Top of the limitation.
 And this can be divided into the machine learning algorithm and the quality of the data. Both of the machine learning algorithm and the quality of the data have decided that the prediction model of the machine learning.

 And now we’re going to talk about the whole process about machine learning. 

Firstly, is the data pretreatment.

 And many according to our experience, we know that it costs about 60 or 70% of the total time.
 The first step is the labels plus the raw data. Which from my point of view that the label is can be determined or can be captured from the raw data or the common knowledge of our human beings. And the raw data come from different types of the source of data.
 And the second step is we have a whole bunch of the data will split them into two kinds of datasets. The first dataset is the training dataset and the second is the test dataset.
 by the meaning of this training, we know that is used for training the model of the machine learning. And for the more, the test data site is used for testing the module which we learned we already get from the training dataset. 

 And more we use the feature extraction and scaling to get the features and get the different scaling.

 In this kind of area, you can use different types of the same. Or the definition, which we called the feature selection and dimensionality reduction. And the sampling as a raw data to the future process.

 And the second one is modeling learning.

  Which we also can be called it is the learning algorithm. In this process, model election, the cross-validation and performance metrics may be the most commonly mentioned. And the haper-parameter optimization is the one we often talk about in this sector. 

As I know that in the modeling learning we want to try different modeling to adjust the raw data and based on our experience or the history called of all the day tomorrow, we want to find out which one is the best based on our experience.

 The 3rd one is the module evaluation.

 Compared to the last two steps. The 3rd step, which we call the evaluation of the model, is easier because you just want to find out which of the model is more useful for us to try the new data or to adjust to the new data, which we can find the data in the test dataset

 And the 4th one is the new sampling prediction.

 According to the definition of this, you can know that the new sampling will be offered by the real world all the different type of data source. We use the source or we can call the raw data to use the algorithm. We learn from the modeling. And we use this model to try to predict the new samples of the data to see what we can use this to predict more hard choices in our real world.