Follow

Understanding and Using the ADAPA Console


The ADAPA Console is your gateway to ADAPA. It allows for the managing and execution of many predictive models. In the Console, a data file can be uploaded and processed against any previously uploaded models. In addition, there is no limit to the number of records that can be scored at any single time, i.e. a data file can be uploaded and processed in its entirety. The resulting file is then available for downloading with the ADAPA computed value for every record (with the exception of records with missing values for which there is no missing value handling strategy defined in the PMML file).

Model results can also be obtained in real-time by means of web-service calls. The help function in the console itself contains a link to documentation on how to have your web-services set-up. The help page also contains a link to a document describing all the data mining techniques supported by ADAPA with a description of their PMML representation. This is useful in case you want to learn how to represent your models in PMML (or find out how they are represented in case you are using an automatically exported PMML file).

If you are using ADAPA on the Cloud, each running ADAPA instance executes its own ADAPA. Therefore, each instance executes its own version of the ADAPA Console. If you have, for example, three instances running on the cloud, you will have access to three different consoles, one for each instance. So, if you want to execute the same model in all instances, you will need to upload the model in all three instances.

 

To sign-up for ADAPA on the Amazon Cloud, simply go to the AWS Marketplace.
 

Using the ADAPA Console

 

The ADAPA Console provides a very easy way to manage models and data in your ADAPA instance. To access all of the functionality offered by the console, you will need to log in first.

1) Providing a Password

 

See our posting on how to log in into the ADAPA Console.

 

2) Manage Models Panel

 

Once you log in, the console will show the "Manage Models" panel (see figure below). This panel is divided into two parts. The first part is represented by a button which allows you to upload your PMML files. The second part, just below the "Upload PMML files" button shows a table containing all the models that have already been uploaded into ADAPA. Note that in the example shown in the figure below, three models have already been uploaded: "AuditSVMModel", "IrisMLRModel", and "LoanNNModel".

Screen_Shot_2014-07-15_at_2.31.54_PM.png
2.1) Uploading PMML Files

To upload your PMML file, simply click on the "Upload PMML Files" button. The "Upload Files" pop-up window will appear (see below). In this window, you can upload one or more files at once. If you are trying to upload multiple PMML files at the same time and you get a "Transfer error" message, make sure to disable the selection of multiple files. When done with file upload, click on "Close". Once you do that, the new file(s) will automatically appear on the table of models.

NOTE: you can also upload a compressed PMML file into ADAPA (.zip). This feature is useful for large model ensembles such as Random Forest Models.

Screen_Shot_2014-07-15_at_2.32.07_PM.png

2.2) Table of Models
The table of available models contains four columns. "Name," "Description," and "Creation Date" are populated with values present in the PMML file itself. If the creation date is missing in your PMML file, then the table will show the time of model uploading. The column "Actions" contains two icons, the first allows you to download the PMML file that has been processed by ADAPA and the second allows you to remove the respective model from ADAPA.
You can also download the PMML file after it has been converted and corrected by ADAPA. You can do that by clicking on the flag next to the model name. In case the flag is yellow or red, as is the case for model "AuditSVMModel", you can download the annotated PMML file. This will contain warnings and errors represented as comments interspersed along the PMML code.

3) Score/Classify Data Panel


Once you click on the bottom panel: "Score/Classify Data", it automatically opens. The figure below shows this panel with the first model from the table of models already selected for scoring. You select another model by clicking on the "Select Model" drop-down menu.

Screen_Shot_2014-07-15_at_2.35.41_PM.png

Once the right model has been selected for scoring, you can choose to batch-score/classify data through it by uploading a data file in CSV format. Files can be uploaded here for model verification or simply for scoring/classification. If your data file contains an expected value for the predicted variable, ADAPA will automatically try to match its computed value with the given expected value. It will then pinpoint any mismatches and give you the opportunity to survey its decisions so that the source of the mismatches can be found (for more on this check postings on model testing).

The figure below shows the data table containing the scoring results after the CSV file was uploaded and processed.

Screen_Shot_2014-07-15_at_2.38.18_PM.png

Note that ADAPA was able to match computed to expected values for all processed records. This is shown on the top part of the data table. This part also contains three buttons: 1) "Download Result": allows you to download the data file you uploaded previously but now with an extra column containing the computed value for every data record; 2) "Settings": allows you to manage the contents of the results table; and 3) "Enable Filters": allows you to search for missing, non-missing, or specific values in the data table.

By clicking on "Settings", the "Result Display Settings" pop-up window will appear.

Screen_Shot_2012-03-26_at_11.37.01_AM.png

Note that it allows you to add any of the input variables to the data table. All you need to do is select a variable from the list and click on the "Show" button. The pop-up window also allows you to change how many records are displayed, the precision of the computed value, and the choice to see only potential mismatched records. This option makes sense only in the case you have uploaded a CSV file for scoring with expected values for model verification.

The figure below shows the result of clicking on button "Enable Filters". Note that once you do that, it allows you to filter records to be displayed depending on different criteria. For example, you can choose to see records starting with "1" for variable "Adjusted". The result of this action is shown in the figure below.

Screen_Shot_2014-07-15_at_2.40.10_PM.png

Simply click on "Remove Filters" to remove the filters from the data table.

Finally, note that on the top part of the ADAPA Console window, there are a set of hyper-links. The first allows you to "Change Password". The "Help" hyper-link gives access to the help page which contains links to documents on how to set up your web-services so that you can access your models in real-time as well as a document describing all the predictive modeling techniques supported by ADAPA and their expression in PMML (see figure below for the help page contents). Use the last hyper-link to "Logout" from the ADAPA Console.
Screen_Shot_2014-07-15_at_2.42.29_PM.png

0 Comments

Article is closed for comments.
Powered by Zendesk