Preparing Data

In this example, the dataset is from the Machine Learning Repository of UCI. For details about the dataset, see Bank Marketing Data Set. Table 1 and Table 2 describe the parameters and sample data of the dataset. You can obtain the dataset from GitHub and upload it to OBS.

1. Download the ModelArts-Lab project from Gitee and obtain the train.csv training data file from the \ModelArts-Lab-master\official_examples\Using_ModelArts_to_Create_a_Bank_Marketing_Application\data directory of the project.
2. Upload the train.csv file to OBS, for example, to the test-modelarts/bank-marketing directory. For details about how to upload files to OBS, see Uploading a File.

Creating a Predictive Analytics Project

1. On the ModelArts management console, click ExeML in the left navigation pane.
2. On the ExeML page, click Create Project in the Predictive Analytics area.
3. On the Create Predictive Analytics Project page, set the project name and select the OBS path where the training data is stored. In this example, the dataset path is test-modelarts/bank-marketing/train.csv. Click Create Project. The data labeling page is displayed, as shown in Figure 1.
   Figure 1 Creating a predictive analytics project
   

Training a Model

1. On the ExeML > Label Data page, preview the data and select the training objective. The training objective here is to determine whether the customer will apply for a deposit (that is, attr_7). Set Label Column Data Type to Discrete Value. After the training objective is specified, click Training.
   Figure 2 Selecting the training objective
   
2. In the displayed Training Configuration dialog box, select an instance flavor used for training and click Yes to start model training.

   The training takes a certain period of time. If you close or exit the page, the system continues training until it is completed.

   Figure 3 Training configuration
   
3. In the upper left corner of the model training page, if the status of the training job changes to Completed, the training job is completed. The Training Details area on the right shows the details about the training job.
   Figure 4 Training job completed
   

Deploying the Model

1. On the Train Model page, click Deploy in the Version Manager area. The system starts to deploy the service and automatically switches to the Deployment Online tab page.
2. In the Version Manager area, when the status changes to Running, the service has been deployed. You can test the service in the prediction area.

   The following shows the test code. As shown in Figure 5, the prediction result is "predictioncol": "yes", indicating that the customer will apply for a deposit.

   {
     "data": 
     {
       "count": 1,
       "req_data": 
   	[
         {
           "attr_1": "58",
           "attr_2": "management",
           "attr_3": "married",
           "attr_4": "tertiary",
           "attr_5": "yes",
           "attr_6": "no",
           "attr_7": ""
         }
       ]
     }
   }

   Figure 5 Testing the model