GB2363216A

GB2363216A - Publish/subscribe data processing with subscriptions based on changing business concepts

Info

Publication number: GB2363216A
Application number: GB0005430A
Authority: GB
Inventors: John Michael Knapman; Robert William Phippen
Original assignee: International Business Machines Corp
Current assignee: International Business Machines Corp
Priority date: 2000-03-08
Filing date: 2000-03-08
Publication date: 2001-12-12
Also published as: AU7197100A; JP2001291044A; GB0005430D0

Abstract

A method comprising steps of: accepting data messages from a publisher application 1; and routing data messages to a subscriber application 3 which has previously registered a subscription request by specifying a name of a topic requesting that published data messages on the topic be routed to the subscriber application; wherein data messages on a first topic are generated for publication by performing data mining data processing operations on data stored in a database 24.

Description

2363216 PUBLISH/SUBSCRIBE DATA PROCESSING WITH SUBSCRIPTIONS BASED ON

CHANGING BUSINESS CONCEPTS

5 Field of the Invention

The present invention relates to the field of data processing and more specifically to event notification data processing which distributes event messages from suppliers (called, hereinafter, "publishers") of data 10 messages to consumers (called, hereinafter "subscribers") of such messages.

While there are many different types of known event notification systems, the subsequent discussion will describe the publish/subscribe event notification system as it is one of the most common.

15 Bac!Mround of the Invention Publish/subscribe data processing systems (and event notification systems in general) have become very popular in recent years as a way of distributing data messages (events) from publishing computers to 20 subscribing computers. The increasing popularity of the Internet, which has connected a wide variety of computers all over the world, has helped to make such publ i shl subscribe systems even more popular. Using the Internet, a World Wide Web browser application (the term "application" or "process" refers to a software program, or portion thereof, running on a computer) 25 can be used in conjunction with the publisher or subscriber in order to graphically display messages. Such systems are especially useful where data supplied by a publisher is constantly changing and a large number of subscribers needs to be quickly updated with the latest data. Perhaps the best example of where this is useful is in the distribution of stock market 30 data.

In such systems, publisher applications of data messages do not need to know the identity or location of the subscriber applications which will receive the messages. The publishers need only connect to a 35 publish/subscribe distribution agent process, which is included in a group of such processes making up a broker network, and send messages to the distribution agent process, specifying the subject of the message to the distribution agent process. The distribution agent process then distributes the published messages to subscriber applications which have previously indicated to the broker network that they would like to receive data messages on particular subjects (topics). Thus, the subscribers also 5 do not need to know the identity or location of the publishers. The subscribers need only connect to a distribution agent process.

Publish/subscribe systems depend on the publisher applications assigning the right data messages to each topic upon which that publisher 10 is publishing, because subscribers subscribe to a given topic expecting to receive messages on that topic. Some topics, however, involve complex processing in order to determine whether a data message should be assigned to a particular topic, the complex operations potentially requiring a large amount of data to be analyzed before a determination for topic assignment 15 can be made. Known publish/subscribe systems have not adequately addresed this, since they assume that publishers can easily assign data messages to topics.

Further, it would be very advantageous to keep the name of each topic 20 the same over time, because this allows subscribers to enter a subscription request once and to not have to repeat the subscription request. However, the precise definition of which messages should be published on a particular topic often is changing on an almost daily basis as business processes evolve and customer behavior changes.

Summary of theInvention

According to a first aspect, the present invention provides an apparatus comprising: means for accepting data messages from a publisher 30 application; and means for routing data messages to a subscriber application which has previously registered a subscription request with the apparatus by specifying a name of a topic requesting that published data messages on the topic be routed to the subscriber application; wherein data messages on a first topic are generated for publication by performing data 35 mining data processing operations on data stored in a database.

According to a second aspect, the present invention provides a method comprising steps of: accepting data messages from a publisher application; and routing data messages to a subscriber application which has Previously registered a subscription request by specifying a name of a topic requesting that published data messages on the topic be routed to the subscriber application; wherein data messages on a first topic are generated for publication by performing data mining data processing 5 operations on data stored in a database.

According to a third aspect, the present invention provides a computer program product stored on a computer readable storage medium for,

when run on a computersystem, instructing the computer system to carry out 10 the method of the second aspect.

In the invention, data mining takes place in order to decide which data messages should be published on a specific topic. This provides a very powerful technique for deciding which messages should be published on 15 a particular topic, especially when large amounts of data and complex data models must be analyzed in order to make such a determination.

Further, with the present invention, the names of the topics can remain the same even though the rules for deciding whether a data message 20 will be published on that topic may change. For example, a publisher may be publishing on a topic called profitable customersff. The precise definition of which customers are profitable may change on an almost daily basis as business processes evolve and customer behavior changes. However,

even though this precise definition is dynamically changin(J, the present 25 invention allows the same topic name to be used by subscribers, and the subscribers will still receive published data messages on the topic named "profitable customers". That is, the content of the subscription will dynamically change while the name of the topic (e.g., profitable customers") remains the same.

Brief Description of the Drawings

The present invention will be better understood by reference to the 35 detailed description of preferred embodiments which will be provided below in conjunction with the following drawing figures:

Fig. 1 is a block diagram showing a message broker implementation according to a preferred embodiment of the present invention; 1, Fig. 2 is a block diagram showing details of the data mining node 22 of Fig. 1; Fig. 3 is a block diagram showing a messaging server implementation according to a second embodiment of the present invention; Fig. 4 is a block diagram showing details of a data mining model according to the preferred embodiment; and Fig. 5 is a block diagram showing a message broker implementation according to a third embodiment of the present invention.

15 Detailed Description of the Preferred Embodiments

In Fig. 1 a first preferred embodiment of the present invention will be described in the context of a multi-function message broker system.

Publish/subscribe broker systems have commonly been integrated into 20 multi-function message broker systems which are used to inter-connect applications which may be on heterogeneous platforms and may use different message formats. For example, Saga Software of Reston, Virginia (USA) have such a message broker product called "Sagavistaff (a trademark of Saga Software). Further, Tibco Software Inc. of Palo Alto, California (USA) 25 also have such a message broker called "TIB/Message Broker" (both "TIB" and "TIB/Message Broker" are trademarks of Tibco). In these multi-function message brokers, a set of pluggable data processing nodes is provided, with each node being dedicated to a specific data processing task, such as message format transformation (e.g., node 21 in Fig. 1), publish/subscribe 30 message distribution (node 23 in Fig. 1), and a rules engine (e.g., 51 in Fig. 5) for deciding (based on a plurality of predefined rules) where an incoming message should be routed.

In these multi-function message broker systems, when a subscriber 35 application 3 registers a subscription request with the broker 2, the subscriber application 3 sends the subscription request to a publish/subscribe broker node 23 specifying the topic of the desired subscription. The publish/subscribe broker node 23 (usually in cooperation with a plurality of other such publish/subscribe broker nodes) then ensures that any published messages on that topic are sent to the subscriber application 3.

In Fig. 1, a publisher application I is publishing messages on a 5 broad general topic called "customers" and provides customer details regarding a plurality of customers of a large business that is making use of the message broker 2. A subscriber application 3 has previously registered a subscription request on the topic "profitable customers" indicating a desire by subscriber 3 to receive published messages 10 concerning customers which have been determined to be profitable. The subscriber application 3 is, for example, the marketing department of a large business that is using the message broker 2 in a customer relationship management (CRM) setting. The incoming messages from publisher 1 are received by the message transformation node 21 where, for is example, the amounts of money associated with a customer's spending are transformed from British pounds to US dollars (we assume in this example that the publisher application I is located in the United Kingdom and maintains customer data based on local British standards and that the message broker is being run in the United States of America). For example,

20 the publisher application 1 could be a British-based provider of customer information.

The received messages are then received by a data mining node 22 which, as shown in Fig. 2, has both a data mining component 221 and a topic 25 assignor component 222. The data mining component 221 takes the customer details included in the messages and performs data mining operations on data stozed in a database 24 by correlating the customer data in the messages with the data stored in the database using well known data mining operations. For example, IBM's Intelligent Miner for Data (TM) is a 30 comprehensive suite of data mining tools which can be used.

As a result of having performed the data mining operations, the topic assignor component 222 of the data mining node 22 makes a determinatlon as to whether a received data message should be assigned to a topic called 35 "profitable customers". Specifically, if the customer details contained in.

the received message have been determined via the data mining component 221 to relate to a "profitable customer" then the topic assignor component 222 will assign the message to the topic "profitable customer". The subscriber 3 will then be provided with the message via the publish/subscribe node 23 (because subscriber 3 has an ongoing subscription to the topic "profitable customer").

Should the large organization running the message broker 2 decide 5 that different criteria should be used for determining whether a customer is profitable, the data mining models used by the data mining component 221 are modified to reflect the changed criteria. Accordingly, an incoming customer details message which previously might have related to a profitable customer will now be found not to relate to a profitable 10 customer. The subscriber 3 need not make any changes to his subscription, as the changed criteria is implemented without affecting pre-existing subscription topics. That is, the subscription topic name stays the same despite the change of criteria as to what will constitute a "profitable customer".

is other typical topics that might be used are:

1) various demographic groups within a customer set 2) credit risks 20 3) complaint letters In a second embodiment, rather than the data mining taking place via one of the nodes in a multifunction message broker, the data mining takes place within a CRM decision server 31 which receives customer details 25 messages from a client application 30 (see Fig. 3). That is, a trained data mining model is deployed as a component of a CRM decision server. The server 31 could act as a subscriber to the messages containing the customer information (e.g., thus making client application 30 a publisher application) if a message broker were inserted in between client 30 application 30 and server 31 of Fig. 3. Otherwise, the client 30 and server 31 can communicate directly using any well-known client/server communication technique (e.g., remote procedure call). Server 31 then performs the data mining operations and sends replies to the client application 30 with the business concept label "profitable customer" 35 matched with each customer detail message for which the server 31 has determined via data mining that the business concept label "profitable customer" should apply. The client application 30 then publishes messages on the topic "profitable customer" to a message broker so that the messages can be received by subscribers who have previously registered subscriptions on the topic "profitable customer".

one of the techniques in IBM's Intelligent Miner for Data (TM) 5 product is a decision-tree classifier. This technique is particularly compatible with a message broker since the predictive models of the decision-tree classifier are in the form of "if-then" rules. In a third embodiment of the present invention, rules of this form are loaded directly into the subscription rule-set of the message broker, enabling the broker 10 to make intelligent decisions about message distribution based on business logic derived from training on suitable business data. See Fig. S where the "if-thenO rules are loaded into a rules processing node 51 which is a standard node of a multi-function message broker. In this third embodiment the rules processing node 51 uses the data mining rules to access the 15 database 24. This should be contrasted with the first embodiment where a dedicated data mining node is provided for carrying out the dating mining operations.

A specific usage scenario, illustrating how a company might make use 20 of the invention, will now be described to illustrate the basic concepts involved with the present invention.

Usage scenario: Customer loyalty Scenario 25 A company which deals directly with customers (either online, or via a call center) wants to augment their service to customers by offering them products which they are likely to want - and consequently improve their loyalty' (likelihood of buying again). They are aware that their market is evolving very quickly, but have a number of key business concepts and 30 descriptions which they want to maintain in the midst of change. The preferred embodiment of the present invention helps to achieve this by separating the definition of the business concept from the label given to it. Consumers of the information (i.e., the various departments of the company) simply have to specify the label' and are then guaranteed to 35 receive information which concerns the current definition of the business concept.

Roles involved; Market analyst Data mining analyst Message broker administrator 5 marketing department, stock control Steps in the process 1. Market analyst: decides that 'Customer loyalty, is an issue and 10 wishes to target 'Affluent, potentially disloyal customers, in an attempt to make them more loyal. The Market analyst knows that a single strategy will not work with all the different types of customer, and thus also wants to build a demographic, customer segmentation to help decide how best to communicate with each different group.

2. Data mining analyst: jointly decides with Market analyst what data features define the terms 'affluent, and loyal,, and builds an initial classification model (41 in Fig. 4) which categorises each customer according to their value to the business and according to their degree of 20 loyalty as a customer. In addition the Data Mining analyst constructs a customer segmentation model 42 which indicates the characteristics of each group. The two models 41 and 42 make up the overall data mining model 40.

These models are very useful in customer relationship management 25 (CRM) applications. In this example, the company is interested in understanding the customer lifetime value, which is the expected value (in terms of profit) which the company can reasonably expect to get from an existing customer. It is often the case that the existing value of a customer (i.e., their total profit to date) is loosely tied to their 30 expected lifetime value. A clear example of this is when banks attempt to recruit university students as customers, at a time when their income and profitability is low, in the expectation that they will eventually become high earners. This example shows how an individual's demographics (in this case, the fact that they are students) can determine their expected value.

35 This principle is used as the basis of data-driven CRM. Some customer activities which affect their value are as follows:

a) "defection" (also known as lapse [insurance industry], churn ftelecoms industry] or attrition [retail industry]: when a previously loyal customer stops buying (and probably switches to another supplier).

5 b) up-selling" is the term used when a customer agrees to purchase a more expensive version of some product they already have.

c) "cross -selling", when a customer buys some product complementary to a purchase they have already made d) "recruitment" is the original acquisition of a (preferably loyal and profitable) customer; e) "fraud" occurs when a customer deceives the company illegally to gain 15 some financial benefit which, from the point of view of the business, causes profitability to become negative.

There are specific techniques which can be used to provide the information needed to manage each of the customer activities mentioned 20 above:

"customer segmentation" divides the customer base into its natural demographic groupings 25 0 "propensity modelling" relates the demographics of a customer to their likelihood of doing something (e.g., buying, committing fraud) "association finding" discovers which pairs of products are likely to be bought together "sequential pattern matching" discovers what sequence of products are commonly bought (e.g., computer, then software, then memory) The IBM Intelligent Miner for Data (TM) data mining software product 35 mentioned above is a comprehensive set of data mining tools which can be applied to each of these issues.

1 10 3. Message broker application administrator: Agrees a 'concept namespace, with the market analyst and the data mining analyst, which is available for any authorised department in the company to view. The business concept list itself can be published and all departments with an interest in this kind of information can be kept up to date with the list of business concepts currently in use (it is expected that this list will change much more slowly than the detailed definition of each concept) 4a. Option 1: Data mining analyst imports the definitions of each classification and segmentation into a 'market analysis server' which is a messaging server. It responds to messages containing customer details by replying with a message containing the business labels for each relevant business concept.

is 4b. Option 2: Message broker application administrator tailors a special message augmentation node by importing the classification and segmentation definitions.

5. Interested departments such as the marketing department and stock 20 control subscribe to business concept labels (topics) which are important to them. For example, the marketing department could subscribe to the topic "affluent" and thus be made aware of customers who are determined (by the latest version of the data mining model) to be affluent customers.

25 6. At a later time, the market analyst and the data mining analyst agree that the current classification for some of the business concepts is outdated. They repeat most of steps 1 and 2. All subscribing/authorised departments continue to receive information about the business concept they specified, and need not know that its detailed definition has changed 30 (though, of course, alerts could optionally be published which give summary details of the change to the concept).

Unless some fundamentally new business concept arises, for example, a new customer segment is recognised (A recent example with Internet shopping 35 is the very new 'Silver Surfer' customer segment; retired people who use the Internet and are prepared to buy online), then no update is required to the business concept labels.

e 11 The present invention is preferably embodied as a computer program product for use with a computer system. Such an implementation may comprise a series of computer readable instructions either fixed on a tangible medium, such as a computer readable media, e.g., diskette, CDROM, 5 ROM, or hard disk, or transmittable to a computer system, via a modem or other interface device, over either a tangible medium, including but not limited to optical or analog communications lines, or intangibly using wireless techniques, including but not limited to microwave, infrared or other transmission techniques. The series of computer readable 10 instructions embodies all or part of the functionality previously described herein.

Those skilled in the art will appreciate that such computer readable instructions can be written in a number of programming languages for use 15 with many computer architectures or operating systems. Purther, such instructions may be stored using any memory technology, present or future,

including but not limited to, semiconductor, magnetic, or optical, or transmitted using any communications technology, present or future, including but not limited to optical, infrared, or microwave. It is 20 contemplated that such a computer program product may be distributed as a removable media with accompanying printed or electronic documentation, e.g., shrink wrapped software, pre-loaded with a computer system, e.g., on a system ROM or fixed disk, or distributed from a server or electronic bulletin board over a network, e.g., the Internet or World Wide Web.

1 12

Claims

1. An apparatus comprising:

means for accepting data messages from a publisher application; and means for routing data messages to a subscriber application which has previously registered a subscription request with the apparatus by 10 specifying a name of a topic requesting that published data messages on the topic be routed to the subscriber application; wherein data messages on a first topic are generated for publication by performing data mining data processing operations on data stored in a database.

2. The apparatus of claim 1 wherein data mining models used in performing the data mining data processing operations are modified in order to change which data messages are generated for publication on the first 20 topic without requiring a change in the name of the first topic.

3. The apparatus of claim 2 wherein the data mining models include a classification model which creates categories of customers of a business which is using the apparatus.

4. The apparatus of claim 3 wherein the data mining models include a segmentation model which includes characteristics of each category of customer.

30 5. The apparatus of claim 1 further comprising:

means for receiving a data message; means for performing data mining operations on data stored in a 35 database based on data included in the received data message; and means for determining whether the data message should be associated with the first topic based on the means for performing.

6. The apparatus of claim 5 wherein the received data message includes customer details regarding a particular customer and the means for performing performs data mining operations on data stored in a database taking the customer details as an input.

7. The apparatus of claim 5 wherein the means for receiving involves subscribing to a second topic which is more general than the first topic.

8. The apparatus of. claim 7 wherein the second topic is customers and 10 the first topic is profitable customers.

9. The apparatus of claim 7 wherein the second topic is customers and the first topic is credit risk customers.

15 10. The apparatus of claim I wherein the data mining data processing operations are performed by a data mining node of a multi-function message broker.

11. The apparatus of claim 1 wherein the data mining data processing 20 operations are performed by a customer relationship management (CRM) server.

12. The apparatus of claim 11 wherein the CRM server is a subscriber to customer information from a client application.

13. The apparatus of claim 1 wherein the data mining data processing operations are performed by a rules processing node of a multi-function message broker.

30 14. A method comprising steps of:

accepting data messages from a publisher application; and routing data messages to a subscriber application which has 35 previously registered a subscription request by specifying a name of a topic requesting that published data messages on the topic be routed to the subscriber application; wherein data messages on a first topic are generated for publication by performing data mining data processing operations on data stored in a database.

5 is. The method of claim 14 wherein data mining models used in performing the data mining data processing operations are modified in order to change which data messages are generated for publication on the first topic without requiring a change in the name of the first topic.

16. The method of claim 15 wherein the data mining models include a classification model which creates categories of customers of a business which is using the apparatus.

17. The method of claim 16 wherein the data mining models include a segmentation model which includes characteristics of each category of customer.

18. The method of claim 14 further comprising steps of:

20 receiving a data message; performing data mining operations on data stored in a database based on data included in the received data message; and 25 determining whether the data message should be associated with the first topic based on the performing step 19. The method of claim 18 wherein the received data message includes customer details regarding a particular customer and the performing step 30 performs data mining operations on data stored in a database taking the customer details as an input.

20. The method of claim 18 wherein the receiving step involves subscribing to a second topic which is more general than the first topic.

21. The method of claim 20 wherein the second topic is customers and the first topic is profitable customers.

7 22. The method of claim 20 wherein the second topic is customers and the first topic is credit risk customers.

23. The method of claim 14 wherein the data mining data processing 5 operations are performed by a data mining node of a multi-function message broker.

24. The method of claim 14 wherein the data mining data processing operations are performed by a customer relationship management (CRM) 10 server.

25. The method of claim 24 wherein the CRM server is a subscriber to customer information from a client application.

15 26. The method of claim 14 wherein the data mining data processing operations are performed by a rules processing node of a multi-function message broker.

27. A computer program product stored on a computer readable storage 20 medium for, when run on a computer system, instructing the computer system to carry out the method of any preceding claim.