Microsoft Azure cloud service introduces the machine learning module. Users only need to upload data and use some algorithm interfaces and R or other language interfaces provided by the machine learning module, you can use Microsoft Azure's powerful cloud computing capabilities to implement your machine learning tasks.
This document introduces the machine learning module and provides a simple application example based on the official examples and help documents. The procedure is as follows.
1. Create a workspace
Note that you must enter a valid Windows Live account when entering the workspace owner.
Enter the created workspace, as shown in figure
2. Upload data
Data source: http://archive.ics.uci.edu/ml/datasets/Statlog+ (German + credit + data)
Download German. Data, a machine learning algorithm used for credit risk. The data includes 20 variables and 1000 credit records, of which 700 are low-risk questions and 300 are high-risk questions. Note that because azure machine learning stidio only supports CSV files, you need to convert German. Data to CSV files.
Click the "+ new" Link under ml studio to set up the data in the workspace as shown in.
3. Create an azure ml Experiment
Click the "+ new" Link under ml studio and select the experiment option.
Step 1: Add a title for this experiment. This article is named "experiment by Jiahua"
Step 2: Find the uploaded data on the left. The name is "UCI German credit card data". Drag the data to the intermediate workspace, the data description is displayed on the right. After the data enters the workspace, it is represented by a rectangle with rounded corners. There is a circle under the rectangle, called "output port". You can place the cursor over it and right-click it to perform data visualization and other operations. Drag the circle to point to the next data processing operation.
Step 3: After adding a dataset, you need to process the dataset, including data preprocessing, Division of training samples and test samples, and selection of machine learning algorithms, detailed operation courses are offered on official instances. After completing the preceding operations, a visual machine learning process is completed, as shown in:
Step 4: run the model. After completing the preceding operations, you can run the program. Click "run" at the bottom to run the model. After each module is run, a green check box is displayed in the upper right corner, if an error occurs in each module or step, a red icon will appear in the same place. After you move the mouse over it, an error type will be displayed.
Step 5: view the result. Right-click the dot in the "Evaluate Model" box and select "Visualize" to view the model running result, as shown in some results:
In this way, a machine learning instance using azure cloud services is completed. Of course, if necessary, you can release the model to the Web server. For details, refer to the help documentation.
Using azure cloud services for machine learning research, my biggest experience is that the visual operation steps make the algorithm running process clearer and clearer, the algorithm is split into data preparation, data preprocessing, training data and test data segmentation, model selection, model parameter adjustment, and model evaluation. Each step is displayed in a processing box, through the line with arrows, we can clearly see the input and output relationships of each link, so that the researchers can easily grasp and control the key points of the algorithm.
This is the end of the preliminary attempt. Due to the relationship between time and capability, this article only makes the simplest attempt. Many details are omitted in the narration process, especially when creating an experiment, for more information, see the official help documentation.