EP1725981A1 - Procede et dispositif de prevision pour evaluer et prevoir des evenements stochastiques - Google Patents
Procede et dispositif de prevision pour evaluer et prevoir des evenements stochastiquesInfo
- Publication number
- EP1725981A1 EP1725981A1 EP05733627A EP05733627A EP1725981A1 EP 1725981 A1 EP1725981 A1 EP 1725981A1 EP 05733627 A EP05733627 A EP 05733627A EP 05733627 A EP05733627 A EP 05733627A EP 1725981 A1 EP1725981 A1 EP 1725981A1
- Authority
- EP
- European Patent Office
- Prior art keywords
- evaluation
- forecasting
- unit
- event
- event data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
Definitions
- the invention relates to a prediction method and apparatus for evaluating and predicting stochastic events.
- Data Mining provides management with insights and relationships that have been hidden or disregarded until now, either because they were considered non-decision-making or un-analyzable.
- rule-based systems are used, which serve to extract the known If-Then rules and to verify them if necessary. Which method is used in each case in the context of "data mining" depends on the respective problem and the area of application.Neuronal networks and systems of linear regression are used in particular for questions with prognosis character.Of course, combinations of the known data mining solutions are conceivable which is usually determined empirically, which date-mining solution is the best method for which application.
- Remote diagnosis and forecasting method for complex systems in particular in connection with vehicle telematics systems, wherein on the basis of recorded on board a vehicle operating data, which are transmitted to a central diagnostic center and thus a remote monitoring is realized, but also a prognosis, such as the probability of failure of individual Determine components.
- a method for predicting a parameter representing the state of a system in particular a
- a very specific technical application of such forecasting method represents the prediction of the operating behavior of a turbine system according to the German patent DE 44 24 743 C2.
- a plant-specific plant model based on the specification of one or more operating parameters further operating parameters determined and taking into account the desired boundary condition or operating parameters Response of the modeled turbines would be calculated on the desired boundary condition and based on which the behavior of the monitored operating parameter or the turbine plant is predicted.
- the invention is therefore based on the object of specifying a prediction method and a prediction device for evaluating and predicting stochastic events, which reacts dynamically to changing boundary conditions and is designed to be self-adaptive.
- an event data record is first created for a process unit which is answered with a binary event value, which is then forwarded to the downstream evaluation unit, whose evaluation result in turn is fed back to another input of the process unit, there is a feedback between the input variables and the output variables in the sense of a simple control loop, so that these changed input variables lead to changed result values, which are included in the forecasting method by way of feedback.
- the event record could include a description of an offer and a customer record, where the binary event value is a digital representation of a purchase offer to the customer Ja / Nein, so that at the cut-off input, a set value can be set for how many customers which a tender offer is being prepared should also accept this.
- the forecasting method according to the invention is thereby carried out in two separate but linked methods, which are controlled by a process unit and a valuation unit.
- the process unit represents the control center of the forecasting process and thus ensures the timing and control of the forecasting process as a whole.
- the process unit has two further outputs for the output of two feature vectors, in which one feature vector comprises the target parameter value, while in the other feature vector the value of the target parameter is still unoccupied. Both feature vectors are then passed to the adjoining valuation unit, which then uses the evaluation of the feature vectors to determine the target parameter value, which is fed back to an additional score input of the processor unit.
- the event data record in the form of an n-tuple at the input of the processor unit, wherein the dimension of the vector is changeable and therefore the value n of the n-tuple is variable.
- the n-tuple does not necessarily have to be normalized. It usually consists of key-value pairs.
- the learning process can be adapted dynamically by self-adaptation of the evaluation system by a simple adaptation of the input data set and / or a changed dimensioning.
- a further significant advantage of the method according to the invention is that the evaluation result of the evaluation unit fed back to the return input of the processor unit is a numerical and thus easily understandable value.
- a high valuation result stands for a high turnover of the customer and a low valuation result for a correspondingly low turnover. This facilitates the practical application the forecasting process considerably.
- the large returned to the return input thus already represents an illustration of a fact.
- the evaluation process taking place in the evaluation unit downstream of the process unit represents a self-adaptive system which has an incremental learning mechanism.
- the method first has to be triggered with predefined training event records which are sequentially applied to the process unit and transferred to the downstream evaluation unit in the form of the feature vectors explained above, with the result that a first optimization of the prognosis method takes place Increasing number of processed event records, especially real event records already an improvement of the system takes place.
- the dynamics of the forecasting method are also made clear by the fact that a temporal evaluation is assigned to the respective evaluation results and, depending on this, a priority weighting.
- the older evaluation results have a lower weight than the younger ones, so that also in this respect changing framework conditions are adequately taken into account.
- This functionality of the evaluation method is suitably described as a "forget function.”
- the dynamics of the prediction method according to the invention are also reflected, inter alia, in an additional setup input of the process unit, via which it is possible to incorporate additional parameters into the current evaluation or the parameters of the event record to be defined, via the setup input that is, the parameterization or predefinition of the event data records created at the input of the process unit. Via the setup input variables can be added on the fly and thus the dimension of the event data record can be extended.
- At least three different process sequences of the prognosis are created within the forecasting process.
- a first method sequence only the event data records are stored in a cache memory assigned to the process unit, and otherwise the number of processed event data sets and the digital evaluation results are paid.
- each customer represented by an incident record is offered a product and returned to each customer as an evaluation result to the return receipt whether the purchase decision was positive.
- the response output 1 is constantly output in this phase.
- the method is ideally implemented in conjunction with a prediction device according to claim 12.
- This forecast device can ideally be operated in conjunction with a conventional data processing system, wherein this data processing system may be in connection with a customer database of a provider, wherein the prediction device is additionally connected to the telephone system of the customer data provider.
- this data processing system may be in connection with a customer database of a provider, wherein the prediction device is additionally connected to the telephone system of the customer data provider.
- the prediction device is additionally connected to the telephone system of the customer data provider.
- Identification features the individual customer data sets are selected, in which case via a display device connected to the forecasting device, a forecast of the expected purchase decisions of the customer with respect to different offers opportunities is displayed.
- a system can be used advantageously in connection with a call center, for example, in which case the respective person in charge is informed of which product or which, depending on the calling customer Service should be offered to the customer in the context of the conversation in order to have an increased probability of making a positive purchase decision.
- FIG. 1 shows a workstation of a call center with connected forecasting device in a block diagram
- FIG. 2 shows a forecasting method of the forecasting device according to FIG. 1 in a block diagram
- FIG. 3 the forecasting method according to FIG. 2 in a detailed block diagram
- FIG. 7 a method for creating so-called scoring cards 8 shows a hydrograph with the course of stochastic events without the use of the prognosis method in comparison to a hydrograph with the use of the prognosis method, each in a diagrammatic representation, with reference to an input variable.
- FIG. 1 shows a typical workstation in a call center. Such a workstation initially consists of the terminal 1 or a computer unit of the respective employee of the call center, which is data-connected to the telecommunications system 2 of the call center. In addition, both the telecommunications system 2 and the terminal 1 with a customer database 3 of the cooperating with the call center provider in data connection.
- the provider can be any participant in the economic process, such as a mail order company, where the call center can either an external or internal device of the provider, which in any case has access to the mentioned customer database 3 of the provider.
- the terminal 1 of the employee of the call center is additionally connected to a forecasting device 4.
- the forecasting device 4 can either be connected to its own display device or use the employee's terminal 1 as a display device.
- an individual offer is then made to the caller, in which case the customer's real purchase decision flows into the customer database 3 and thus flows into the evaluation by the forecasting device 4 on the next call from the same customer.
- Computing unit can be realized, essentially of two mutually connected in data connection blocks.
- the event data sets are transferred to the forecasting device 4. These are vectors, so-called n-tuples, which are applied to a request input 11 of the prediction device 4.
- the response output 12 is followed by a query unit 13, which in the case the event value output at the response output 12 is 0, the event data record created at the request input 11 is discarded in an elimination step 9 or else a further processing is initiated.
- the offer is submitted to the customer, ie an external process 8 is switched on and the customer reaction in the form of a numerical evaluation result is further fed back via the feedback path 7 to a return input 10 of the forecasting device 4. It can simply be the feedback, the customer has bought something, or what is the turnover or something similar.
- the forecasting device 4 is additionally provided with a cut-off input 14, at which the ratio of the digital event values with one another, that is the percentage of the events rated 1 with respect to the total number of event data sets, can be set.
- the forecasting device 4 is parameterized.
- the number of dimensions ie the number n of the n-tuple of the event data sets created at the request input 11
- the parameters of the form and the name contained in the event data record and Type can be defined according to.
- key-value pairs such as "age: 35"
- the forecasting device comprises a process unit 5 with downstream evaluation unit 6.
- the downstream evaluation device 6 has at least two Inputs 16, 17 on. These are a train input 16 and a score input 17, which are each connected to a train output 20 and a request output 21 of the process unit 5.
- the downstream evaluation process is rejected in the evaluation unit 6 as a function of whether an input value 1 or 0 is output at the response output 12, wherein the event value 1 in the present example represents the recommendation to the customer may submit.
- the two inputs 16, 17 of the evaluation unit 6 are each occupied by the feature vectors output by the process unit, wherein the train input 16 for adaptation of the applied in the evaluation unit 6
- Evaluation processes is used and therefore requires a feature vector with an occupied target variable, wherein the score input 17, a feature vector is created, the target variable is not occupied.
- a score result is output at a score output 22.
- the score value output at score output 22 represents a numeric number corresponding to the target variable already mentioned. This target variable or this evaluation value is returned to an additional score input 23 of the process unit 5. The evaluation value determined by the evaluation unit 6 thus flows into the further evaluation by the forecasting device 4.
- each event data record reported to the prediction apparatus 4 is stored in a so-called request cache 24, and as soon as the feedback evaluation values are applied to the return input 10 of the process unit 5, the event vector previously stored in the request cache 24 is retrieved using the data records of the event vector Evaluation unit 6 feedback-enriched value enriched and then the complete record so n-tuple include the event record and the evaluation value in one
- Train cache 25 is stored. As soon as the train cache 25 is full, the evaluation unit 6 is trained with the content of the train cache 25. In the event that the event data set can not be found in the request cache 24 on the basis of the values feedback-coupled by the process unit 5, an error message 26 is output.
- the train cache 25 is assigned a threshold value query 27, via which it is checked in each case whether the train cache 25 has already run to full, that is, a predetermined number of event data sets has been created in this memory element. Once this number is reached, these event records are used, for example, to improve the parameterization of the evaluation unit 6, wherein the model underlying the forecasting device 4 is trained in a training step 30 and then the train cache 25 is emptied in an emptying step 28.
- At least three different process sequences can be distinguished from one another in the inventive prediction method, whereby it depends on the respective learning success and learning progress of the inventive prediction method which of the possible method sequences is used.
- the different ones Processes are reflected in particular in the processing of the applied to the input of the process unit event data sets, the so-called Requests 35.
- the requests 35 are written unchanged in a write step of the process unit associated request cache 24 and increased for each request 35 of the threshold value 31 by one, and the integrated response payer for each request 35, which is answered with the response 1 , is also raised by one.
- this first phase but at
- the internal parameter sets of the evaluation unit 6 are adapted to the changed situation. Now, however, the evaluation unit 6 causes either an event value 1 or 0 at the response output 12 can be output. Therefore, a response query 37 must be switched before the parameter setting 37. It is thus carried out in parallel a rating of how well the forecasting device 4 is already working, so about how often the customer
- Acceptance offer This ratio is again monitored by a threshold value 31.
- a further defined threshold value When a further defined threshold value is reached, a change is made from the second method sequence to a third method sequence which essentially differs from the second method sequence only in that the internal parameters used in the evaluation unit 6 are changed as a function of the learning outcome of the prediction method.
- the prediction apparatus 4 can be preceded by a training database 40 from the simulation unit 41, which is also connected upstream Endless loop 42
- a parallel validation database 43 serve to check the good of the forecasting device 4 with an independent data set, if necessary.
- FIG. are three time-lapse hydrographs in three superimposed diagrams.
- the top diagram shows the progression of the positive purchase decisions, ie the return values 1 in relation to the total number of processed event records.
- a value of 0.5 is set.
- the positive return curve very quickly and noticeably approaches the desired sales success.
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Strategic Management (AREA)
- Finance (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE102004013020A DE102004013020A1 (de) | 2004-03-16 | 2004-03-16 | Prognoseverfahren und -vorrichtung zur Bewertung und Vorhersage stochastischer Ereignisse |
PCT/DE2005/000479 WO2005091183A1 (fr) | 2004-03-16 | 2005-03-16 | Procede et dispositif de prevision pour evaluer et prevoir des evenements stochastiques |
Publications (1)
Publication Number | Publication Date |
---|---|
EP1725981A1 true EP1725981A1 (fr) | 2006-11-29 |
Family
ID=34964927
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP05733627A Ceased EP1725981A1 (fr) | 2004-03-16 | 2005-03-16 | Procede et dispositif de prevision pour evaluer et prevoir des evenements stochastiques |
Country Status (5)
Country | Link |
---|---|
US (1) | US20080147702A1 (fr) |
EP (1) | EP1725981A1 (fr) |
AU (1) | AU2005224715A1 (fr) |
DE (2) | DE202004021667U1 (fr) |
WO (1) | WO2005091183A1 (fr) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9336493B2 (en) * | 2011-06-06 | 2016-05-10 | Sas Institute Inc. | Systems and methods for clustering time series data based on forecast distributions |
CN113256325A (zh) * | 2021-04-21 | 2021-08-13 | 北京巅峰科技有限公司 | 二手车估价方法、系统、计算设备和存储介质 |
Family Cites Families (25)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5276771A (en) * | 1991-12-27 | 1994-01-04 | R & D Associates | Rapidly converging projective neural network |
DE4424743C2 (de) | 1994-07-13 | 1996-06-20 | Siemens Ag | Verfahren und Vorrichtung zur Diagnose und Prognose des Betriebsverhaltens einer Turbinenanlage |
AU3477397A (en) * | 1996-06-04 | 1998-01-05 | Paul J. Werbos | 3-brain architecture for an intelligent decision and control system |
US5832466A (en) * | 1996-08-12 | 1998-11-03 | International Neural Machines Inc. | System and method for dynamic learning control in genetically enhanced back-propagation neural networks |
DE19753034A1 (de) | 1997-11-18 | 1999-06-17 | Ddg Ges Fuer Verkehrsdaten Mbh | Verfahren zur Prognose eines den Zustand eines Systems repräsentierenden Parameters, insbesondere eines den Zustand eines Verkehrsnetzes repräsentierenden Verkehrsparameters und Vorrichtung zum Durchführen des Verfahrens |
US7912761B2 (en) * | 1999-08-27 | 2011-03-22 | Tech Venture Associates, Inc. | Initial product offering system and method |
US7013285B1 (en) * | 2000-03-29 | 2006-03-14 | Shopzilla, Inc. | System and method for data collection, evaluation, information generation, and presentation |
US6662192B1 (en) * | 2000-03-29 | 2003-12-09 | Bizrate.Com | System and method for data collection, evaluation, information generation, and presentation |
US6539392B1 (en) * | 2000-03-29 | 2003-03-25 | Bizrate.Com | System and method for data collection, evaluation, information generation, and presentation |
US20020059154A1 (en) * | 2000-04-24 | 2002-05-16 | Rodvold David M. | Method for simultaneously optimizing artificial neural network inputs and architectures using genetic algorithms |
US7742959B2 (en) * | 2000-05-01 | 2010-06-22 | Mueller Ulrich A | Filtering of high frequency time series data |
US6954758B1 (en) * | 2000-06-30 | 2005-10-11 | Ncr Corporation | Building predictive models within interactive business analysis processes |
AU2002227514A1 (en) * | 2000-07-27 | 2002-02-13 | Polygnostics Limited | Collaborative filtering |
US20060271441A1 (en) * | 2000-11-14 | 2006-11-30 | Mueller Raymond J | Method and apparatus for dynamic rule and/or offer generation |
US6954757B2 (en) * | 2001-02-02 | 2005-10-11 | Hewlett-Packard Development Company, L.P. | Framework, architecture, method and system for reducing latency of business operations of an enterprise |
GB2373347B (en) * | 2001-03-07 | 2006-11-22 | Touch Clarity Ltd | Control system to actuate a robotic operating system |
GB0122121D0 (en) * | 2001-09-13 | 2001-10-31 | Koninkl Philips Electronics Nv | Edge termination in a trench-gate mosfet |
EP1493113A4 (fr) * | 2002-03-20 | 2009-04-22 | Catalina Marketing Corp | Stimulations ciblees se basant sur un comportement predit |
US6745151B2 (en) * | 2002-05-16 | 2004-06-01 | Ford Global Technologies, Llc | Remote diagnostics and prognostics methods for complex systems |
US20030229884A1 (en) * | 2002-05-21 | 2003-12-11 | Hewlett-Packard Development Company | Interaction manager template |
US20030220860A1 (en) * | 2002-05-24 | 2003-11-27 | Hewlett-Packard Development Company,L.P. | Knowledge discovery through an analytic learning cycle |
CA2436400A1 (fr) * | 2002-07-30 | 2004-01-30 | Abel G. Wolman | Geometrisation servant a la reconnaissance des formes, l'analyse des donnees, la fusion des donnees et la prise de decisions a plusieurs criteres |
AU2003272981A1 (en) * | 2002-10-10 | 2004-05-04 | Sony Corporation | Robot device operation control device and operation control method |
EP1586076A2 (fr) * | 2003-01-15 | 2005-10-19 | Bracco Imaging S.p.A. | Systeme et procede d'optimisation d'une base de donnees pour l'entrainement et le test d'algorithmes de prediction |
US20050154701A1 (en) * | 2003-12-01 | 2005-07-14 | Parunak H. Van D. | Dynamic information extraction with self-organizing evidence construction |
-
2004
- 2004-03-16 DE DE202004021667U patent/DE202004021667U1/de not_active Expired - Lifetime
- 2004-03-16 DE DE102004013020A patent/DE102004013020A1/de not_active Withdrawn
-
2005
- 2005-03-16 US US10/592,731 patent/US20080147702A1/en not_active Abandoned
- 2005-03-16 WO PCT/DE2005/000479 patent/WO2005091183A1/fr active Application Filing
- 2005-03-16 AU AU2005224715A patent/AU2005224715A1/en not_active Abandoned
- 2005-03-16 EP EP05733627A patent/EP1725981A1/fr not_active Ceased
Non-Patent Citations (2)
Title |
---|
None * |
See also references of WO2005091183A1 * |
Also Published As
Publication number | Publication date |
---|---|
DE202004021667U1 (de) | 2010-05-12 |
AU2005224715A1 (en) | 2005-09-29 |
DE102004013020A1 (de) | 2005-10-06 |
US20080147702A1 (en) | 2008-06-19 |
WO2005091183A1 (fr) | 2005-09-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
DE102012009502A1 (de) | Verfahren zum Trainieren eines künstlichen neuronalen Netzes | |
DE102006002247A1 (de) | Verfahren und System zum Aktualisieren von Echtzeitdaten zwischen Intervallen | |
DE69730130T2 (de) | Betrugsüberwachung in einem fernmeldenetz | |
EP2879582B1 (fr) | Dispositif, procédé et application de détermination d'un niveau de charge actuel | |
EP1472630A2 (fr) | Procede pour la reconnaissance automatique de formes et pour la regulation de processus economiques ainsi que systeme de reconnaissance de formes et autopilote pour des processus economiques avec prise en consideration de la valeur de rendement des relations-clients | |
DE112021000370T5 (de) | Auf maschinellem lernen beruhende datenüberwachung | |
DE4203233A1 (de) | Verfahren und vorrichtung zum loesen von problemen | |
DE69821933T2 (de) | System und verfahren zum erstellen von erfüllungsregeln für handelsysteme | |
AT412678B (de) | Verfahren zur rechnergestützten erstellung von prognosen für operative systeme sowie system zur erstellung von prognosen für operative systeme | |
DE102010038930A1 (de) | Verfahren und System zum Erzeugen eines Prognosenetzwerkes | |
EP1725981A1 (fr) | Procede et dispositif de prevision pour evaluer et prevoir des evenements stochastiques | |
DE202020104077U1 (de) | Strategiegenerierungssystem für die automatische Auswertung und den automatischen Handel | |
EP3812949A1 (fr) | Jumeaux numériques configurables | |
DE602004006648T2 (de) | Benutzungsverfahren und system in einem kommunikationsnetz | |
EP1081665B1 (fr) | Système expert | |
DE202022104425U1 (de) | Intelligentes System für die sichere Integration von Kreditwürdigkeitsprüfungen und Bankensystemen durch maschinelles Lernen | |
EP0956531A1 (fr) | Procede pour la transformation d'une logique floue servant a la simulation d'un processus technique en un reseau neuronal | |
EP1208541B1 (fr) | Systeme expert | |
EP3716058A1 (fr) | Procédé de commande d'un appareil au moyen d'un nouveau code de programme | |
EP4016202B1 (fr) | Procédé de reproduction des composantes de bruit des signaux de fonctionnement enregistrés avec pertes et dispositif de commande | |
AT501028B1 (de) | Verfahren und vorrichtung zur bewertung von börsekursen | |
DE10061108A1 (de) | Verfahren zur Vorhersage von Börsenkursen | |
EP3913566A1 (fr) | Système d'investissement ainsi que procédé d'évaluation | |
DE102020205687A1 (de) | Verfahren und Vorrichtung zum Prädizieren einer Entwicklung von Krankheitsfällen einer infektiösen Krankheit | |
DE102021211830A1 (de) | Verfahren für eine Überprüfung einer komplexen Regulationssituation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
17P | Request for examination filed |
Effective date: 20060823 |
|
AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IS IT LI LT LU MC NL PL PT RO SE SI SK TR |
|
17Q | First examination report despatched |
Effective date: 20070122 |
|
RAP1 | Party data changed (applicant data changed or rights of an application transferred) |
Owner name: EPOQ GMBH |
|
DAX | Request for extension of the european patent (deleted) | ||
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R003 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE APPLICATION HAS BEEN REFUSED |
|
18R | Application refused |
Effective date: 20110319 |