DE19705614A1 - Fully automated advisory system for customers - Google Patents

Abstract

The expert system has a parameter data memory 2 and a program memory 3and these connect with a n inference machine 4 having a processor 5 and a second program memory 6. Inputs to the inference machine are provided by a customer specific receiver 8 via a telecommunication link 12,12' coupled to a data bases 9,9'.

Description

The invention relates to an expert system with at least one egg Data storage for parameter data, with a first pro gram memory for storing process sequences and with an inference machine that has a processor and a connected second program memory for playback a processing program, the Infe process machine from the first program cher together with corresponding parameter data from the Da memory by means of the processing program on gangs data applied via an input to the expert system and conclusions from it Form of output data, which is calculated via an output can be queried.  

Such expert systems are known, for example, from the textbook by Chadwick and Hannah entitled "EXPERT SYSTEMS FOR MICROCOMPUTERS ", Sigma Press (1987), pages 1 to 7.

In such expert systems, an "artificial Intel ligenz "created by making experience values more human "Experts" either as machining processes or rules in a computer program in the inference machine input fin the and taking into account boundary conditions that as Parameter data also contain empirical values, the Sy enable the system to process current input data as a human expert would do and as a result nis conclusions and instructions to be drawn in the form of initial data can be taken from the expert system nen.

Known expert systems usually exist physically from a computer with a processor on its processor with corresponding procedures and another Program memory connected to a processing program sen, and has access to a data storage corresponding parameter data for various boundary conditions scenarios investigated with the expert system should be. The user of the expert system gives the answer appropriate input data to solve a current problem for example in the form of inquiries and receives after corresponding operation of the inference engine in response Instructions that show him how the current one Problem could be solved. The input person is in direct spatial contact with which the expert system containing calculator. Communication with the Expertensy stem over long distances is not yet planned.  

On the other hand, online services already exist which can be queried remotely, but which only the transmission of current data without the possibility offer "intelligent" processing. In particular There is no way to query problem solving over larger spatial distances.

In contrast, the object of the present invention is a Fully automated database system for advice, for example wise from customers even over long distances on the base an expert system of the type described at the beginning create, depending on customer requirements customer-specific offers can also be created, for example can, with completely different data qualities for Processing can find its way into the expert system, and being on changes in the market or in the area of supply ters can be reacted to immediately.

According to the invention this object is achieved in that on Input a first receiving device is provided which customer-specific input data of at least two under received different data sources and to the Inferencema machine can forward, at least one of the data sources are spatially separated from the expert system, and that the parameter data in the data memory via a second Receiving device are changeable, the provider-specific can receive specific and / or market-specific data.

This enables a customer to use a telecommunications, for example network from afar with the expert system contact record and its custom data along with an inquiry or request. At the same time  the system, possibly also from a distance, for example se via databases or current information services receive the latest market-specific data from which in Ex expert system suitable current boundary conditions for the asked problem solving can be formed. Another, completely different quality can be provider-specific Have data that is also remotely, but also in the Close range can be entered in the receiving device nen. It is also conceivable that, for example, a bank customer ne customer-specific data in the near range, while the market-specific data, for example from a far distant stock exchange or from remotely located to provide certain services through telecommunications networks can be entered.

An embodiment of the invention is particularly preferred according expert system, in which the first program memory or a third pro connected to the inference machine the structure of a neural network points. Such neural networks are known per se. They have a logical structure like that in the brain more sophisticated animals, which leads to that these systems are learnable and depend on certain procedures runs can be pre-trained. It is extremely flexible Adaptation to current, changing problem over time structures possible.

An embodiment of the invention is also particularly preferred expert system according to the invention, in which the first program memory or a third connected to the inference machine Program memory based on fuzzy logic contains. Such fuzzy logics are also known per se. They include "fuzzy" rules that stare towards each other  grammatical constraints, as otherwise described in Compu programs are usual, a significantly increased Fle Open flexibility for the approximate solution of problems. In particular, a problem formulation in the form of natural language possible when using fuzzy logic.

At least one of the data sources preferably has one considerable spatial distance, namely at least 10 m preferably over 1 km to the expert system, so that the Expert system according to the invention without problems, for example in a telebanking system can be integrated.

An embodiment is particularly preferred in which the via the second receiving device into the data memory incoming parameter data in two or more sub-databases ken can be stored for each incoming data channel. Different time constants from require different data qualities considering fin the.

In a further preferred embodiment of the invent Expert systems according to the invention are the first and / or second Receiving device and preferably also the output on or several telecommunication networks connected. On the In this way, a particularly wide range of users of the Systems can be achieved.

A further development of this embodiment is particularly advantageous form in which the expert system in an on-line service system is integrated. Such on-line services are today already widespread and usually particularly user-friendly friendly.  

Ver. Also falls within the scope of the present invention using an expert system of the type described above for automatic provision of financial services. Next the use of a falls within the scope of the invention Expert system for automatic Bear described above processing applications to authorities or public services put.

Further advantages of the invention result from the Be writing and drawing. Likewise, the above mentioned and invented the features further invented in accordance with the invention individually for themselves or for several in be arbitrary combinations are used. The shown and The embodiments described are not intended to be final Understand enumeration, but rather have exemplary th character for the description of the invention.

The invention is illustrated in the drawing and is hand explained in more detail by exemplary embodiments. Show it:

Figure 1 is a schematic system overview of the effect of the expert system according to the invention. and

Fig. 2 is a diagram of the internal structure of the expert system according to the invention.

Algorithms have been used in the field since the early 1990s artificial neural networks, fuzzy logic and the genetic or evolutionary algorithms into one Class of procedures called Computational In telligence, biological or soft computing or Intel ligent information processing. Surrender by combining the different processes in  hybrid systems new fields of application. In addition to technical Applications are primarily applications from business and Administration now widespread and currently come from the Universities in the application. Strengths of the new techniques are the creation of forecasts (e.g. share prices, Business development), the rule-based or experience-based Decision support or the determination of white starting factors such as creditworthiness or risk from a lot number of input parameters.

The expert system according to the invention is an integrated system system for fully automatic advice, calculation, risky determination up to the preparation of offers in the investment, real estate lien, credit, insurance and bancassurance in Self-service operation via online services.

FIG. 1 shows a schematic overview of the functioning of the expert system according to the invention. On the one hand, provider-specific data and market-specific data are entered into the system, on the other hand, customer-specific data over longer distances, for example with the help of telecommunications networks and online services. The expert system then performs an automatic quotation calculation and generation and is sufficient The finished offer can be visualized back to the customer via the corresponding online service. In addition, automated advice can also be given to the customer, who in turn can place an order via the same data line if the offer meets the customer's wishes.

Through this self-service principle, the provider can enor save me personnel costs. In consulting-intensive industries can at least be a preparation for later counseling speak within and outside normal opening hours of the provider in an automatic way through the expert system occur.

Another advantage of the system is that everyone time an automatic consideration of changes the influencing factors on the provider, market or customer side is possible.

Finally, the system according to the invention can also be used as an inte Integrated component for automatic needs analysis and an bidding used in non-requested areas will.

The following table schematically shows the application examples for the expert system according to the invention from the areas of vehicle insurance, mortgage loans, stock pot management and property purchase:

In general there is the technical solution of the invention according to integrated expert systems from the following components ten:

Customer system (e.g. standard PC with modem) with components for

• - network connection
• - Visualization of the information transmitted
• - Entry and encryption of personal data
• - Device for the inclusion of the access authorization need data.

Supplier system with components for

• - Collection and storage of market-specific data (e.g. database with loss statistics, share prices, real estate bilien,. . .). Data is e.g. B. also via network be updated from time to time and serve as a calculation basis for the offer.
• - Collection and storage of provider-specific data, e.g. B. Requirements for profit margins, commissions, tariffs,. . .
• - Collection and storage of customer-specific data, e.g. B. vehicle class, creditworthiness,. . . (see table). Possibly. Risi Co or credit rating from this data with the help of fuzzy logic by linking individual criteria.
• - Expert system based on saved experience evaluate and by processing the customer, market and provider-specific data with systems capable of learning (e.g. neural networks) and / or based on rules (fuzzy Logic) decisions or offer data determined. Opti Male systems can consist of hybrid neuro / fuzzy components being constructed.

Security component:

• - Encryption device to protect the transmitted Data
• - If necessary with access control to regulate access calculations (password, PIN, face recognition via video, Fingerprint, speaker verification,. . .)  
• - electronic signature
• - Automatic fraud detection using typical Transaction patterns

Network component:

• - for transmission z. B. the calculated offer or Decision on lending through the network the user.
• - to transfer the encrypted information to both Directions

Extension for combined services:

• - Second expert system for automatic detection of a further customer needs (via the current request ) by target group determination with intelligent In formation processing methods, e.g. B. rule-based with Fuzzy logic. Example: Additional for car insurance inquiries automatic creation of a life insurance offer if the following (possibly weighted) criteria partially or at all predominantly fulfilled: age between 18 and 55, permanent residence in the EU, vehicle type at least middle class or second car and regular Income available.

A typical procedure, for example in the area of real estate transactions, can look like this:

• 1. Someone wants to buy a house.  
• 2. It communicates the desired properties of the house to a database (e.g. number of rooms, living area, approximate location of the property, price, additional costs Etc.)
• 3. The service provider stores corresponding information (Interest, level of interest, economic situation on).
• 4. The customer activates the database and transmits his Wishes (number of rooms; area where the house is located should; maximum purchase price; maximum debt burden that the Customer can afford to buy a house can).
• 5. The database creates an offer about available houses with the features you want and the best type of fi financing and transmits the offer to the customer.

The logical structure of an embodiment of the expert system according to the invention is shown schematically in FIG. 2:

The expert system 1 has a data memory 2 for parameter data which can be entered via a second receiving device 11 and updated as desired. In particular, the entered parameter data can originate from different sources, with a separate sub-database then being set up for each incoming data channel in the data memory 2 , on which the corresponding parameter data are stored.

Furthermore, the expert system 1 contains a first program memory 3 , which in the exemplary embodiment shown can have the structure of a learnable, pre-trainable neural network.

The data memory 2 and the first program memory 3 are connected to the heart of the expert system 1 , which is formed by an inference machine 4 . This comprises a processor 5 and a second program memory 6 connected thereto, which contains a processing program for the operation of the processor 5 .

Furthermore, in the embodiment shown in FIG. 2, a third program memory 10 is contained in the expert system 1 , which is also connected to the inference machine 4 . In the exemplary embodiment shown, this third program memory 10 can contain “unsharp” rules that are based on fuzzy logic.

The parameter data received in the data storage 2 via the second receiving device 11 can be composed of both market-specific and provider-specific data and can be entered either spatially directly or over greater distances, for example via a telecommunications network. The input can be made as required, regularly, irregularly or continuously. So as market-specific data such. B. Continuously receive market values from stock exchanges from all over the world, while the provider-specific data entry only takes place when the business strategy changes.

The parameter data are particularly effective in neural networks especially on the weighting factors at the neural link network structure, while in a fuzzy lo gik directly influence or change the fuzzy rules.

Via a first receiving device 8 , customer-specific input data from different data sources 9 , 9 ′, 9 ″ are entered into the expert system 1 , the data source 9, for example, representing a telecommunication terminal of a spatially distant customer who makes a request together with the can enter customer-specific data via a telecommunications network 12 into the receiving device 8 of the expert system 1 . Furthermore, for example via another telecommunication network 12 ', data from another customer can be entered into the receiving device 8 from a data source 9 ' which is also remote. Finally, walk-in customers can enter customer-specific data into the expert system 1 via a further data source 9 ″, for example a customer terminal, which is immediately adjacent to one another.

After appropriate processing of the incoming data, taking into account boundary conditions, rules and experience-based procedures using the neuronal network in the first program memory 3 , the inference machine 4 creates a result of the complex problem, which is in the form of answers to an output 7 of the expert system 1 given and from there, for example, can be returned to the customer via the telecommunications network 12 .

Claims (9)

1. Expert system ( 1 ) with at least one data memory ( 2 ) for parameter data, with a first program memory ( 3 ) for storing method sequences and with an inference machine ( 4 ) which has a processor ( 5 ) and a second program memory ( 6 ) for storing a processing program, whereby the inference machine ( 5 ) applies process sequences from the first program memory ( 3 ) together with corresponding parameter data from the data memory ( 2 ) by means of the processing program to input data which are input to the expert system ( 1 ) are entered, and from this conclusions are calculated in the form of output data which can be queried via an output ( 7 ), characterized in that a first receiving device ( 8 ) is provided at the input, the customer-specific input data from at least two different data sources ( 9 , 9 ', 9 '') received and sent to the Can forward inference machine ( 5 ), at least one of the data sources ( 9 , 9 ') being spatially separated from the expert system, and that the parameter data in the data memory ( 2 ) can be changed via a second receiving device ( 11 ), which is provider-specific and / or can receive market-specific data. 2. Expert system according to claim 1, characterized in that the first program memory ( 3 ) or a third with the inference machine ( 5 ) connected program memory ( 10 ) has the structure of a neural network. 3. Expert system according to claim 1 or 2, characterized in that the first program memory ( 3 ) or a third program memory ( 10 ) connected to the inference machine ( 5 ) contains program rules based on fuzzy logic. 4. Expert system according to one of the preceding claims, characterized in that at least one of the data sources ( 9 , 9 ') has a spatial distance of at least 10 m, preferably over 1 km to the expert system ( 1 ). 5. Expert system according to one of the preceding claims, characterized in that the via the second receiving device ( 11 ) in the data memory ( 2 ) incoming parameter data can be stored in two or more sub-databases for each incoming data channel. 6. Expert system according to one of the preceding claims, characterized in that the first and / or second receiving device ( 8 or 11 ) and preferably also the output ( 7 ) to one or more telecommunications networks ( 12 , 12 ') are connected. 7. Expert system according to one of the preceding claims, preferably according to claim 6, characterized in that the expert system ( 1 ) is integrated in an on-line service system. 8. Use of the expert system according to one of the previous claims for the automatic provision of Fi financial services. 9. Use of the expert system according to one of the claims 1 to 7 for the automatic processing of applications at Authorities or public services.