2
The OP Administrative UI

The OP Administrative UI is a browser-based user interface that permits the OP administrator to manage RE Farms and REs, schedule and deploy packages, produce reports, and otherwise manage OP.

This chapter starts with an overview of the Administrative UI and then illustrates how to use the UI to obtain recommendations.

Administrative UI Overview

Bring up the OP Administrative UI by typing the following in the URL field of your browser:

   http://<hostname>/OP/Admin/

where <hostname> is the name of the system on which Oracle9iAS is installed.

Note that the URL is case-sensitive.

The first page that appears is the login page, which welcomes you to Oracle9iAS Personalization.

Login Page

Enter user name and password for the MOR schema (a valid user name and password were established during OP installation; if you don't know what they are, ask the person who installed OP).

• After entering a valid user name and password, click Log in, which brings up the OP Administrative home page.

Home Page

The Home page welcomes you and briefly describes the product. At the left are links to common tasks, such as create new farm, create package, schedule a build, etc. They are listed in the order you would follow in executing the tasks. To the right is a brief status of recent events (builds, deployments, and reports).

Browse the pages to get familiar with their structure and content. Click the tabs to see the various pages. The tabs are Home, Farms, Packages, Schedules, Reports, and Log. The tab pages organize OP administrative tasks.

Farms Page

The Farms page lists the current recommendation engine farms. Use the Farms page to create, delete, and in general manage recommendation engine farms.

The Farms page gives access to recommendation engines, which you add to a farm. Adding a recommendation engine to a farm means specifying the RE's database connection details. These details were established during the installation of the database.

An RE must have a connection to the MTR. If there is no pre-existing MTR connection, you create one and give it a name. This also requires information established during installation and configuration of the database.

Packages Page

The Packages page lists current packages. Use this page to create, delete, and manage packages.

An OP package contains all the information needed for building predictive models, which includes the general settings for the package, information about its connection to the MTR, and settings specified for the package's build (which includes settings for building the model).

To create a package, click Create Package. This brings up the first of three pages that guide you through creating a package:

• The first page, Create Package, prompts you for a name and description of the package and asks you to specify the MTR connection by name. If there is no MTR connection, you are propted to create one.

• The second page, Specify Build Settings, prompts you for the build settings, which include
  • Recommendation performance: This is a trade-off between accuracy and speed -- that is, between higher recommendations accuracy and faster recommendations.

  • New products: Specify whether you want OP to use a proxy for new products. There must be proxy information in the MTR to use this option.

  • Choose what time-stamped session data to use: Use data from any dates, including data with no dates; a rolling window of the past x number of days, or between specified dates.

• The third page, Confirm Settings, summarizes and confirms the settings you have specified.

The Confirm Settings page also gives you the option of going directly to scheduling a build for the package.

Schedules Page

Click the Schedules tab and note the three kinds of events that are scheduled: Builds, deployments, and reports. The Schedules tab opens by default to the Build Schedule page (see the next section).

Build Schedule

Use this page to create, edit, delete, and in general manage the building schedule of a package, that is, the creation of predictive models and other information needed to make recommendations.

To create a new build schedule, click Create Build Schedule. The Create Build Schedule page gives you the following options:

• Build the package as soon as possible.

• Build the package at a future time that you specify, and with the frequency you specify. When a package is built, it is stored in the MOR, ready for deployment.

This page also gives you the option of going directly to scheduling deployment of the package (see the next section).

Deployment Schedule

Use this page to create, edit, delete, and in general manage deployment schedules.

Deploying a package means copying it to every RE on a given RE Farm. Multiple REs with the same package can share the load.

To create a new deployment schedule, click Create Deployment Schedule. The Create Deployment Schedule page gives you the following options:

• Deploy the package every time it is built.

• Deploy the package as soon as possible.

• Deploy the package at a future time that you specify, and with the frequency you specify.

Report Schedule

Use this page to create and manage report schedules.

To schedule a report, click Create report schedule.

The Create Report Schedule page gives you the following options:

• General settings: Specify the RE Farm, the report type, the email address of the person who should receive notification, and any notes you may wish to make.

• Data selection: Use data from any dates; a rolling window of the past x number of days, or between specified dates.

• Schedule: Run the report as soon as possible or schedule the run for a specified date with specified frequency.

The Create Report Schedule page also lets you do the following:

• To update or look at the details of a report schedule, click its Edit icon.

• To delete a scheduled report, select it and click Delete.

• To stop a scheduled report that is running, select it and click Stop.

Note: You cannot edit or delete a scheduled report while it is running or canceling; it must be idle.

Reports Page

Use this page to view reports. There are three types of reports:

• Purchasing session report: Reports, for a user-specified period of time, the total number of sessions, the number of sessions that had at least one purchase, and the percentage of purchasing sessions to total sessions.

  The total number of sessions includes both visitor and customer sessions. Sessions that had purchases include only customer sessions.

• Recommendation effectiveness report: Reports, for a user-specified time period, the number of recommendation requests served by the system, the number of recommended items clicked per recommendation request, and the number of recommended items purchased per recommendation request.

  This report includes all clicked and purchased items in its "Clicked" and "Purchased" columns, even those items that are not part of the recommendations.

• Itemized recommendation effectiveness report: Reports, for a specified time period, how many times the items have been recommended, how many times they have been clicked, and how many times they have been purchased.

  Entries in the "Product ID" column have the following format:

  <Product type> : <Product id>

Note: The MTR tables MTR_SESSION and MTR_RECOMMENDATION_DETAIL must be populated before a report can be generated.

Log Page

The event log allows you to monitor results of scheduled builds, deployments, and reports.

To view details of an item, click its Details icon. To delete one or more items, select the item(s) and click Delete.

Obtaining OP Recommendations

You must deploy a package before you can execute a Java program that requests recommendations or collects data.

Therefore, before you can obtain recommendations you must:

1. Create a Recommendation Engine Farm.

2. Add at least one RE to the Farm.

3. Populate at least the minimum data required in the MTR tables. You can do this with the demo data that is loaded at the time of installataion.

4. Schedule a build and deployment.

This chapter illustrates performing these steps.

Note that you cannot build a package unless there is data available; see "Data for Model Building" in Chapter 1 for more information.

Create an RE Farm with an RE

The first step is to create an RE Farm. There are two ways to start:

• On the Home page, click the Create new farm link to bring up the Create Recommendation Engine Farm page.

• Click the Farms tab to go to the Farms page. On the Farms page, click Create Farm (lower right), which brings up the Create Recommendation Engine Farm page.

On the Create Recommendation Engine Farm page,

1. Enter a name for the farm.

2. Click Add Recommendation Engine.

3. On the Add Recommendation Engine page, enter a name for the recommendation engine.

   For the database connection details, you will need information that was provided during installation.

   • For Host ID, Port, SID, and Database alias, you will need the database information you provided when you installed the database.

   • For DB schema name, User name, and Password, you will need the information provided when you installed and configured the RE schema.

4. After filling in all the fields, click Test Connection to determine whether the database connection is successful, that is, to verify that the information provided can be used to make a successful connection to the desired schema.

5. Assuming the database connection is successful, click OK (lower right), which takes you back to the Create Recommendation Engine Farm page, where you will see listed the recommendation engine you defined.

   Note: If you click Cancel instead of OK, the information you have entered is lost.

Create a Package

Next, create a package.

To create a package, you must have a connection to the MTR. If you do not have an MTR connection,

1. Click Options (upper right) and go to the MTR database connections section. The information needed for an MTR connection comes from entries provided when you installed and configured the MTR schema.

2. To check the database connection, click Test Connection.

3. Assuming the connection is successful, click OK, which returns you to the Create Farm page.

Now you can create a package:

1. Click the Packages tab, which brings up the Packages page.

2. On the Packages page, click Create Package (lower right), which brings up the first of three pages of the Create Package wizard.

3. Give the package a name (required), a description (optional), confirm the name of the MTR connection, and click Next.

4. Specify the settings to be used to build and tune the new package, and click Next.

5. Review the settings you have specified, leave the Schedule a build box checked, and click Finish. This takes you to the Create Build Schedule page.

Schedule a Build

On the Create Build Schedule page, select

1. Build as soon as possible.

2. Leave the Schedule deployment box checked.

3. Click OK. This takes you to the Create Deployment Schedule page.

Schedule a Deployment

On the Create Deployment Schedule page,

1. Select Deploy every time the package is built.

2. Leave the default for Frequency, Once.

3. Click OK. This returns you to the Packages page.

Summary

You have created an RE Farm with one RE, and you have created a package and scheduled its build and deployment.

Check the Packages page after a few minutes to see whether the package has yet built and deployed. The Packages page dispays the status; refresh the page by clicking Go.

When the package has built and deployed successfully, you can use it to collect data and make recommendations using the Recommendation Engine API.

MTR Sample

Next, browse the contents of the Mining Table Repository (MTR) database used to build the model. This is the prepopulated MTR that is installed when OP is installed if you select that option. This prepopulated MTR provides the data needed to perform the exercises described in this manual.

Use SQL*Plus commands to examine the contents of any of the database tables. The table below shows what part of one of the MTR database tables looks like. It contains movie ratings by customers, demographic data on those customers, an ID for each movie that was rated, the rating given the movie by the customer, and the data source type.

CUSTOMER_ID	ITEM_ID	ITEM_TYPE	ATTRIBUTE_ID	BIN_VALUE	DATA_SOURCE_TYPE
2	264	MOVIE	1	1	2
2	389	MOVIE	2	1	3
2	153	MOVIE	2	2	3
2	354	MOVIE	2	2	3
2	264	MOVIE	2	3	3
2	153	MOVIE	3	1	4
2	264	MOVIE	3	1	4
2	354	MOVIE	3	1	4
2	389	MOVIE	3	1	4
2	0	NONE	4	3	1
2	0	NONE	5	1	1
2	0	NONE	6	2	1

This sample comes from a large database table that contains movie ratings by customers, demographic data on those customers, an ID for each movie that was rated, and the rating given the movie by the customer. Table columns are as follows:

• CUSTOMER_ID: The data is all from customer with CUSTOMER_ID = 2.

• ITEM_ID: The ID numbers identify particular movies. Demographic information for a customer always has an ITEM_ID of 0.

• ITEM_TYPE: Refers to kind of item being rated; here the type is movie. Demographic informtion for a customer always has an ITEM_TYPE of NONE.

• ATTRIBUTE_ID: Values 1, 2, and 3 are attributes from purchasing, rating, and navigational data; values 4, 5, and 6 are from demographic data (the last three rows), where 4 = age, 5 = gender, and 6 = marital status.

• BIN_VALUE: Ratings were binned in 3 bins; the values here correspond to those bins.

• DATA_SOURCE_TYPE: (what kind of data) 1 = demographic, 2 = purchases, 3 = ratings, 4 = navigational.

For more information about the OP schemas, see the Oracle9iAS Personalization Administrator's Guide.