WO2021242223A1 - Predictive and interventive intelligence - Google Patents

Predictive and interventive intelligence Download PDF

Info

Publication number
WO2021242223A1
WO2021242223A1 PCT/US2020/034561 US2020034561W WO2021242223A1 WO 2021242223 A1 WO2021242223 A1 WO 2021242223A1 US 2020034561 W US2020034561 W US 2020034561W WO 2021242223 A1 WO2021242223 A1 WO 2021242223A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
parameters
normal
historical data
profiles
Prior art date
Application number
PCT/US2020/034561
Other languages
French (fr)
Inventor
Severence M. MACLAUGHLIN
Original Assignee
Delorean Artificial Intelligence, Llc
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Delorean Artificial Intelligence, Llc filed Critical Delorean Artificial Intelligence, Llc
Priority to MX2022014681A priority Critical patent/MX2022014681A/en
Priority to PCT/US2020/034561 priority patent/WO2021242223A1/en
Priority to US17/999,045 priority patent/US20230177537A1/en
Priority to EP20938380.1A priority patent/EP4154209A4/en
Priority to CN202080101992.0A priority patent/CN116324832A/en
Publication of WO2021242223A1 publication Critical patent/WO2021242223A1/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION 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
    • G06Q30/00Commerce
    • G06Q30/02Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201Market modelling; Market analysis; Collecting market data
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION 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
    • G06Q30/00Commerce
    • G06Q30/02Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201Market modelling; Market analysis; Collecting market data
    • G06Q30/0202Market predictions or forecasting for commercial activities
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/21Design, administration or maintenance of databases
    • G06F16/219Managing data history or versioning
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24Querying
    • G06F16/242Query formulation
    • G06F16/2433Query languages
    • G06F16/244Grouping and aggregation
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION 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
    • G06Q10/00Administration; Management
    • G06Q10/06Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION 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
    • G06Q30/00Commerce
    • G06Q30/01Customer relationship services
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00Machine learning

Definitions

  • the present invention relates generally to data modeling, and more particularly to predictive and interventive intelligence.
  • businesses today are looking for smarter ways to streamline their existing sales operations process and improve conversion rates. For example, businesses look to chase opportunities that have higher changes of conversion and profitability, shorten sales cycles without compromising conversion odds, and reducing cost of winning deals and ongoing operations.
  • the invention features a method including receiving historical data housed in the one or more interconnected computer systems in a first computer system, the historical data comprising structured data and unstructured data, storing the historical data in a central database residing in the first computer system in the network, aggregating the historical data in the central database in the first computer system according to subject matter, validating the aggregated historical data in the first computer system, analyzing the validated aggregated historical data in the first computer system using a series of tools, generating normal profiles in the first computer system from the analyzed validated aggregated historical data, storing the generated normal profiles as horizon data sets, receiving real time data housed in the one or more of the interconnected computer systems in the first computer system, comparing the real time data to the horizon data sets to identify normal and out of normal profiles, and generating rules-based outcomes for out of normal profiles based on parameters for investigation.
  • the invention features a system including a processor, a memory, a data store, the memory including an operating system and a predictive process, the predictive process including receiving historical data housed in one or more interconnected computer systems, the historical data comprising structured data and unstructured data, storing the historical data in the data store, aggregating the historical data in the data store according to subject matter, validating the aggregated historical data, analyzing the validated aggregated historical data using a series of tools, generating normal profiles from the analyzed validated aggregated historical data, storing the generated normal profiles as horizon data sets, receiving real time data from the one or more of the interconnected computer systems, comparing the real time data to the horizon data sets to identify normal and out of normal profiles, and generating rules-based outcomes for out of normal profiles based on parameters for investigation.
  • FIG. 1 is a block diagram of an exemplary network.
  • FIG. 2 is a block diagram of an exemplary server.
  • FIG. 3 is a flow diagram.
  • FIG. 4 is a flow diagram.
  • FIG. 5 illustrates an exemplary CRM framework.
  • FIG. 6 illustrates some exemplary types of data ingested.
  • FIG. 7 illustrates an example of next best action.
  • an exemplary network 10 includes a local area network (LAN) 12 and a link 14 to network 16 of interconnected computers (e.g., Internet).
  • the LAN 12 includes one of more server systems 18, 20 and 22. Although only three servers 18, 20, 22 are shown, the LAN 12 may include any number of servers.
  • the servers 18, 20, 22 are all linked to each other using wired and/or wireless communication.
  • server 18 communicates with the network 16 of interconnected computers over link 14.
  • the link 14 may be wired and/or wireless.
  • Each of the server systems 18, 20, 22 and each of the interconnected computers in the network 16 include at least a processor, a memory and a datastore.
  • the datastore may include a repository of public and/or private information.
  • the server system 18 includes a processor 50, a memory 52 and a storage unit 54.
  • the memory 52 includes at least an operating system (OS) 56, such as Windows®, Unix® or Linux®, and a predictive process 100, fully described below.
  • OS operating system
  • predictive process 100 includes receiving (102) historical data housed in the one or more interconnected computer systems in a first computer system.
  • Receiving (102) may include, for example, acquiring historical data.
  • the interconnected computer systems may include, for example, cloud systems, data marts, data lakes, smart data lakes, data oceans and so forth.
  • the historical data may also include data received from additional computer systems linked to the network.
  • the received data may include, for example, structured data and unstructured data.
  • structured data is data that resides in a fixed field within a record or file.
  • unstructured data is data that does not have a pre-defined data model or is not organized in a pre-defined manner.
  • the received data may include, for example, internal and external data. Further, when the N is large enough 'thick' data can be incorporated. External data can be purchased or "harvested,” meaning that it is scraped from the digitized universe .
  • Predictive process 100 stores (104) the historical data in a central database residing in the first computer system in the network.
  • the historical data may be stored (104) in, for example, a cloud systems, a data mart, a data lake, a smart data lake, a data oceans and so forth.
  • Predictive process 100 aggregates (106) the historical data in the central database in the first computer system.
  • Aggregating (106) permits multiple data sources referencing one subject to be analyzed.
  • Predictive process 100 validates (108) the aggregated historical data in the first computer system.
  • Predictive process 100 analyzes (110) the validated aggregated historical data in the first computer system using a series of analytic, artificial intelligence or data science tools.
  • the analytic, artificial intelligence or data science tools may include but not limited to one of more of text mining, natural language processing (NLP), natural language utilization (NLU), natural language generation (NLG), machine learning (e.g., supervised and unsupervised), genetic programming, evolutionary programming, deep learning, neural networks, computer vision, and audio sound recognition.
  • Predictive process 100 generates (112) normal profiles in the first computer system from the analyzed validated aggregated historical data.
  • the normal profiles represent mathematical models of potential groups that may be defined by mathematical patters, repeatable behaviors, characteristics, pathways, common variables, causal events, risk or return on investment (ROI) clusters, and so forth.
  • the potential groups may be, for example, clients, patients, customers, workers, and so forth.
  • the normal profiles may include sub-profiles, such as, for example, customers in New York, Connecticut and Massachusetts.
  • Predictive process 100 stores (114) storing (114) the generated normal profiles as historical horizon data sets.
  • the historical horizon data sets represent profiles of a target industry, such as, for example, consumers, patients, physicians, salespersons, sales deals, and so forth.
  • Predictive process 100 receives (116) real time data housed in the one or more of the interconnected computer systems in the first computer system.
  • Real time data may be different for various industries. For example, in the financial services trading, real time is milliseconds, for the pharmaceutical industry real time is daily, and for the fashion industry trends, real time is quarterly.
  • real time data is data that is analyzed and quantified to generate insights and predictions based on a need of an industry and or a client and a speed of which data can be acquired and ingested.
  • Predictive process 100 compares (118) the real time data to the historical horizon data sets to identify normal and out of normal profiles.
  • Predictive process 100 generates (120) rules-based and or predictive outcomes for normal profiles and out-of-normal profiles based on risk, financial, operations, quality and other business or industry parameters for investigation.
  • Predictive process 100 introduces (122) a normal or out of normal profiles into a theoretical scenario and uses (124) the introduced normal profiles to predict a probable outcome of the theoretical scenario.
  • the introduced normal profile and/or out of normal profile to predict a probable outcome of the theoretical scenario may include matching mathematical profiles of the normal profile to situational event data and variables of the theoretical scenario.
  • Using (124) may include an intervention phase where best actions lead to mitigation, intervention or enhancement of an event and/or probability of a future interaction.
  • process 100 implements rules-based, predictive and/or interventive outcomes based on risk parameters for investigation, utilizing robotic processing automation (RPA) and/or other technologies and analyses. These profiles can be shunted to either a predetermined outcome based on lower risk or rules or directly to human intervention. Further, if there is something that falls outside of normal, these profiles are modeled for differences and how those differences could impact the nascent industry parameters to predict what outcomes would happen.
  • RPA robotic processing automation
  • Process 100 Determining whether an event in the future will happen with a probability is what process 100 can achieve. Once process 100 has a flow through of real time data that is ever expanding the historical horizon data set as well as real time analysis of that data flow, process 100 has the ability to understand and analyze how the mathematical profiles will interact within different types of situational environments that are constructed with theoretical data variables. Process 100 understands how and what a profile is and how it will interact when presented into a situation. Thus, if a profile is introduced into a theoretical scenario, process 100 can predict its outcome with a known probability. The more situational data and the richer the data is, the higher the precision in the validity of the probability.
  • Process 100 can include attack algorithms, which be analogized to baby hunting dogs. These attack algorithms are raised in a sand box and trained to "hunt" or identify a specific mathematical profile, just as a hunting puppy is trained to learn a smell of prey. Once the attack algorithms are trained, they can be released to the digitized universe and when they identify the mathematical profile that they are trained to recognize, then they red flag identification and trigger a prediction or analyses.
  • attack algorithms which be analogized to baby hunting dogs. These attack algorithms are raised in a sand box and trained to "hunt” or identify a specific mathematical profile, just as a hunting puppy is trained to learn a smell of prey. Once the attack algorithms are trained, they can be released to the digitized universe and when they identify the mathematical profile that they are trained to recognize, then they red flag identification and trigger a prediction or analyses.
  • the attack algorithm is the antibody that has been developed to identify those proteins and latch onto for identification for the immune system.
  • the immune system in this analogy is process 100.
  • the attack algorithm is basically a highly focused and profile centric advanced WebCrawler/DigiCrawler.
  • a business or organization can determine if this event is good or bad and whether it will have positive or negative effects to the organization.
  • the interventive phase of process 100 enables an organization to change the probability that the event will occur. If the event is determined to have negative outcomes, the organization can choose to pull levers or take preventative actions in order to mitigate the effect of this event. If the event is positive to operations, then the organization can take steps to enhance the probability of the event.
  • process 100 matches mathematical profiles of incoming data to situational event data and variables and determines a probability of an event happening in the future, the situational variables are given weight based on how much that variable contributes to the event happening.
  • the interventive phase is all about how does one pull certain levers or how to change these variables to either mitigate or enhance the future event. In data science and artificial intelligence consulting this is referred to as "next best action.” Because process 100 knows what influences the event, process 100 can generate a report detailing what variables are influencing the occurrence and by what weight and prioritize their importance. From this prioritization process 100 can recommend or take actions that will change the outcome of the predicted event.
  • CRM customer relationship management
  • Process 100 may be used mathematically model deals that are present in each of the different phases (A to F) as well as the progression forward and backwards between phases of the sales funnel (G to M). Further, process 100 can characterize and model those deals that fail or drop out of the funnel at specific stages (N to Q) and more powerfully the process 100 can determine what is different from those deals that drop out or lose and those that progress to the next stage and eventually win.
  • predictive process 100 can be used in a number of areas, such as, for example, healthcare, where it can be used in modeling real time medical management, disease progression, claims, payment integrity, and so forth.
  • Predictive process 100 can be used in customer facing, e.g., how to identify a customer, mathematically model who the customer is, how to contact the customer at the right time, in the right channel (e.g., email, phone, mail, app, and so forth) with the right message to drive conversion, and so forth.
  • the right channel e.g., email, phone, mail, app, and so forth
  • Predictive process 100 can be used to acquire and retain the right customers, especially within loyalty programs.
  • Predictive process 100 can be used in life sciences, e.g., patient and prescriber identification.
  • Predictive process 100 can be used in the natural resources and oil and gas industry, e.g., how to identify what fields of natural resources are most profitable and how to develop in the most efficient and profitable manner.
  • Predictive process 100 can be used with commodity, stock and financial instruments, e.g. to predict price and trading value in future.
  • an exemplary CRM framework 200 is illustrated, which depicts a process from data acquisition through four veins of analysis, i.e., structured data, supervised learning, unsupervised learning and sentiment analysis.
  • the result is a deal win/loss probability, where the deal will fail if it does not win and what next best actions can be taken to increase probability of a win.
  • Data is harvested from a backend of a CRM system for both structured and unstructured data.
  • structured data are client demographics, days in deal, days in stage, salesperson, categorization, and so forth.
  • Structured data is anything entered in a specific header or predetermined field for the CRM basic or customizable.
  • Unstructured data includes but not limited to free text notes in specific fields, meeting notes, follow up notes, meeting invites, acceptances and associated emails, emails pertaining to the deal, uploaded documents, uploaded presentations and any other potential files such as PDFTM files, ExcelTM files, and so forth.
  • FIG. 6 illustrates some exemplary types of data ingested. Data is placed in a data base and then analyzed in two veins, i.e., structured and unstructured.
  • Structured data is aggregated and a master data file is created. Each phase or stage is mathematically modeled from horizon data. Data is then enriched from unstructured data. Machine learning is used to determine win/loss prediction classifiers. Machine learning is used to determine probability of deal movement classifiers from one stage to the next and in what direction. The Machine learning results in key deal progression drivers and situational data variables, and what next best action to take. An example of next best action is illustrated in FIG. 7 where situational data variables that are weighted in the prediction of deal event in the future and potential next best actions to improve deal win rate and progression through the sales funnel. Attribution modeling helps uncover how different intervention levers come together to drive/increase in overall deal conversion probabilities or deal TCV (directly or indirectly) at a per-deal or at a client/account level.
  • Unstructured data analysis typical includes three prongs, i.e., unsupervised learning, supervised learning and sentiment analysis.
  • Unstructured data sources such as free text notes, emails and or files are harvested, the unstructured data is then parsed and extracted, data is cleaned and then vectorized. From this point the auto-prepared data is then divided into unsupervised and supervised veins of analysis.
  • the unstructured data is then processed into clusters that are then mathematically modeled per stage and the progression between stages. The result of this process is the ability give a probability based on unstructured text and documents as to what stage in the CRM process the deal is present or transitioning. Once the different stage clusters are mathematically modeled, the unstructured data is analyzed for sentiment analysis.
  • unstructured data that has been cleansed, parsed and vectorized are labeled based in the unsupervised learning results of the clusters and mathematical models of each sales stage and or transitioning.
  • a machine learning neural net model is trained based on segmentation and characterization per sales stage and transition phase, which allows for auto-classification and probability referencing of unstructured data as to what phase the deal is currently assigned.
  • the summation of the unsupervised, supervised and sentiment analysis results are fed into the data enrichment of the structured data for further machine learning modeling.
  • the four veins of analysis, i.e., structured, unsupervised learning, supervised learning and sentiment analysis give final results of:
  • the present invention enables predictive modeling of the progression of a customer / deal along the sales funnel by leveraging both the structured and unstructured data.
  • the present invention identifies the factors controlling the progression of a customer / deal along the sales funnel.
  • the present invention predicts the end state (target) probabilities of a new customer / deal and up to what stage (along the funnel) a new customer / a new deal can progress (if the target state probabilities are not favorable).

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Strategic Management (AREA)
  • Theoretical Computer Science (AREA)
  • Development Economics (AREA)
  • Physics & Mathematics (AREA)
  • Finance (AREA)
  • Accounting & Taxation (AREA)
  • General Physics & Mathematics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Economics (AREA)
  • Data Mining & Analysis (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Databases & Information Systems (AREA)
  • Game Theory and Decision Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Educational Administration (AREA)
  • Tourism & Hospitality (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

A method includes receiving historical data housed in the one or more computer systems, the historical data including structured data and unstructured data, storing the historical data in a central database, aggregating the historical data in the central database according to subject matter, validating the aggregated historical data, analyzing the validated aggregated historical data using a series of tools, generating normal profiles in the first computer system from the analyzed validated aggregated historical data, storing the generated normal profiles as horizon data sets, receiving real time data housed in the one or more computer systems, comparing the real time data to the horizon data sets to identify normal and out of normal profiles, and generating rules-based, predictive and interventive outcomes for out of normal profiles based on parameters for investigation.

Description

PREDICTIVE AND INTERVENTIVE INTELLIGENCE
Applicant:
DeLorean Artificial Intelligence, LLC
BACKGROUND OF THE INVENTION
The present invention relates generally to data modeling, and more particularly to predictive and interventive intelligence.
In general, businesses today are looking for smarter ways to streamline their existing sales operations process and improve conversion rates. For example, businesses look to chase opportunities that have higher changes of conversion and profitability, shorten sales cycles without compromising conversion odds, and reducing cost of winning deals and ongoing operations.
SUMMARY OF THE INVENTION
The following presents a simplified summary of the innovation in order to provide a basic understanding of some aspects of the invention. This summary is not an extensive overview of the invention. It is intended to neither identify key or critical elements of the invention nor delineate the scope of the invention. Its sole purpose is to present some concepts of the invention in a simplified form as a prelude to the more detailed description that is presented later.
In general, in one aspect, the invention features a method including receiving historical data housed in the one or more interconnected computer systems in a first computer system, the historical data comprising structured data and unstructured data, storing the historical data in a central database residing in the first computer system in the network, aggregating the historical data in the central database in the first computer system according to subject matter, validating the aggregated historical data in the first computer system, analyzing the validated aggregated historical data in the first computer system using a series of tools, generating normal profiles in the first computer system from the analyzed validated aggregated historical data, storing the generated normal profiles as horizon data sets, receiving real time data housed in the one or more of the interconnected computer systems in the first computer system, comparing the real time data to the horizon data sets to identify normal and out of normal profiles, and generating rules-based outcomes for out of normal profiles based on parameters for investigation.
In another aspect, the invention features a system including a processor, a memory, a data store, the memory including an operating system and a predictive process, the predictive process including receiving historical data housed in one or more interconnected computer systems, the historical data comprising structured data and unstructured data, storing the historical data in the data store, aggregating the historical data in the data store according to subject matter, validating the aggregated historical data, analyzing the validated aggregated historical data using a series of tools, generating normal profiles from the analyzed validated aggregated historical data, storing the generated normal profiles as horizon data sets, receiving real time data from the one or more of the interconnected computer systems, comparing the real time data to the horizon data sets to identify normal and out of normal profiles, and generating rules-based outcomes for out of normal profiles based on parameters for investigation.
These and other features and advantages will be apparent from a reading of the following detailed description and a review of the associated drawings. It is to be understood that both the foregoing general description and the following detailed description are explanatory only and are not restrictive of aspects as claimed.
BRIEF DESCRIPTION OF THE DRAWINGS
These and other features, aspects, and advantages of the present invention will become better understood with reference to the following description, appended claims, and accompanying drawings where:
FIG. 1 is a block diagram of an exemplary network.
FIG. 2 is a block diagram of an exemplary server.
FIG. 3 is a flow diagram.
FIG. 4 is a flow diagram.
FIG. 5 illustrates an exemplary CRM framework.
FIG. 6 illustrates some exemplary types of data ingested.
FIG. 7 illustrates an example of next best action.
DETAILED DESCRIPTION
The subject innovation is now described with reference to the drawings, wherein like reference numerals are used to refer to like elements throughout. In the following description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the present invention. It may be evident, however, that the present invention may be practiced without these specific details. In other instances, well-known structures and devices are shown in block diagram form in order to facilitate describing the present invention.
As shown in FIG. 1, an exemplary network 10 includes a local area network (LAN) 12 and a link 14 to network 16 of interconnected computers (e.g., Internet). The LAN 12 includes one of more server systems 18, 20 and 22. Although only three servers 18, 20, 22 are shown, the LAN 12 may include any number of servers. The servers 18, 20, 22 are all linked to each other using wired and/or wireless communication. In the example shown, server 18 communicates with the network 16 of interconnected computers over link 14. The link 14 may be wired and/or wireless.
Each of the server systems 18, 20, 22 and each of the interconnected computers in the network 16 include at least a processor, a memory and a datastore. The datastore may include a repository of public and/or private information.
As shown in FIG. 2, the server system 18, for example, includes a processor 50, a memory 52 and a storage unit 54. The memory 52 includes at least an operating system (OS) 56, such as Windows®, Unix® or Linux®, and a predictive process 100, fully described below.
As shown in FIG. 3, predictive process 100 includes receiving (102) historical data housed in the one or more interconnected computer systems in a first computer system. Receiving (102) may include, for example, acquiring historical data. The interconnected computer systems may include, for example, cloud systems, data marts, data lakes, smart data lakes, data oceans and so forth. The historical data may also include data received from additional computer systems linked to the network.
The received data may include, for example, structured data and unstructured data. In general, structured data is data that resides in a fixed field within a record or file. In general, unstructured data is data that does not have a pre-defined data model or is not organized in a pre-defined manner.
The received data may include, for example, internal and external data. Further, when the N is large enough 'thick' data can be incorporated. External data can be purchased or "harvested," meaning that it is scraped from the digitized universe .
Predictive process 100 stores (104) the historical data in a central database residing in the first computer system in the network. In other embodiments, the historical data may be stored (104) in, for example, a cloud systems, a data mart, a data lake, a smart data lake, a data oceans and so forth.
Predictive process 100 aggregates (106) the historical data in the central database in the first computer system.
Aggregating (106) permits multiple data sources referencing one subject to be analyzed.
Predictive process 100 validates (108) the aggregated historical data in the first computer system.
Predictive process 100 analyzes (110) the validated aggregated historical data in the first computer system using a series of analytic, artificial intelligence or data science tools. The analytic, artificial intelligence or data science tools may include but not limited to one of more of text mining, natural language processing (NLP), natural language utilization (NLU), natural language generation (NLG), machine learning (e.g., supervised and unsupervised), genetic programming, evolutionary programming, deep learning, neural networks, computer vision, and audio sound recognition.
Predictive process 100 generates (112) normal profiles in the first computer system from the analyzed validated aggregated historical data. The normal profiles represent mathematical models of potential groups that may be defined by mathematical patters, repeatable behaviors, characteristics, pathways, common variables, causal events, risk or return on investment (ROI) clusters, and so forth. The potential groups may be, for example, clients, patients, customers, workers, and so forth.
The normal profiles may include sub-profiles, such as, for example, customers in New York, Connecticut and Massachusetts.
Predictive process 100 stores (114) storing (114) the generated normal profiles as historical horizon data sets. The historical horizon data sets represent profiles of a target industry, such as, for example, consumers, patients, physicians, salespersons, sales deals, and so forth.
Predictive process 100 receives (116) real time data housed in the one or more of the interconnected computer systems in the first computer system. Real time data may be different for various industries. For example, in the financial services trading, real time is milliseconds, for the pharmaceutical industry real time is daily, and for the fashion industry trends, real time is quarterly. Here we consider real time data to be decision time data. Decision time data is data that is analyzed and quantified to generate insights and predictions based on a need of an industry and or a client and a speed of which data can be acquired and ingested.
Predictive process 100 compares (118) the real time data to the historical horizon data sets to identify normal and out of normal profiles.
Predictive process 100 generates (120) rules-based and or predictive outcomes for normal profiles and out-of-normal profiles based on risk, financial, operations, quality and other business or industry parameters for investigation.
Predictive process 100 introduces (122) a normal or out of normal profiles into a theoretical scenario and uses (124) the introduced normal profiles to predict a probable outcome of the theoretical scenario. Using (124) the introduced normal profile and/or out of normal profile to predict a probable outcome of the theoretical scenario may include matching mathematical profiles of the normal profile to situational event data and variables of the theoretical scenario. Using (124) may include an intervention phase where best actions lead to mitigation, intervention or enhancement of an event and/or probability of a future interaction.
Once process 100 is complete, and one understands the mathematical models of what is normal and what is not normal, one can then use these mathematical models in the historical horizon data set as a "mirror" to compare data that is coming in real/decision time to see if it matches profiles within the normal parameters or whether they fall outside of what is normal. Those profiles that fall outside of the normal mathematical model may then flagged for further analysis. Based on these analyses, process 100 implements rules-based, predictive and/or interventive outcomes based on risk parameters for investigation, utilizing robotic processing automation (RPA) and/or other technologies and analyses. These profiles can be shunted to either a predetermined outcome based on lower risk or rules or directly to human intervention. Further, if there is something that falls outside of normal, these profiles are modeled for differences and how those differences could impact the nascent industry parameters to predict what outcomes would happen.
Determining whether an event in the future will happen with a probability is what process 100 can achieve. Once process 100 has a flow through of real time data that is ever expanding the historical horizon data set as well as real time analysis of that data flow, process 100 has the ability to understand and analyze how the mathematical profiles will interact within different types of situational environments that are constructed with theoretical data variables. Process 100 understands how and what a profile is and how it will interact when presented into a situation. Thus, if a profile is introduced into a theoretical scenario, process 100 can predict its outcome with a known probability. The more situational data and the richer the data is, the higher the precision in the validity of the probability.
Process 100 can include attack algorithms, which be analogized to baby hunting dogs. These attack algorithms are raised in a sand box and trained to "hunt" or identify a specific mathematical profile, just as a hunting puppy is trained to learn a smell of prey. Once the attack algorithms are trained, they can be released to the digitized universe and when they identify the mathematical profile that they are trained to recognize, then they red flag identification and trigger a prediction or analyses.
Another analogy to explain an attack algorithm is to think of a profile as a cell with certain cellular membrane proteins that represent variables with data points. The attack algorithm is the antibody that has been developed to identify those proteins and latch onto for identification for the immune system. The immune system in this analogy is process 100. The attack algorithm is basically a highly focused and profile centric advanced WebCrawler/DigiCrawler.
Once an event is predicted to occur in the future, a business or organization can determine if this event is good or bad and whether it will have positive or negative effects to the organization. The interventive phase of process 100 enables an organization to change the probability that the event will occur. If the event is determined to have negative outcomes, the organization can choose to pull levers or take preventative actions in order to mitigate the effect of this event. If the event is positive to operations, then the organization can take steps to enhance the probability of the event.
When process 100 matches mathematical profiles of incoming data to situational event data and variables and determines a probability of an event happening in the future, the situational variables are given weight based on how much that variable contributes to the event happening. The interventive phase is all about how does one pull certain levers or how to change these variables to either mitigate or enhance the future event. In data science and artificial intelligence consulting this is referred to as "next best action." Because process 100 knows what influences the event, process 100 can generate a report detailing what variables are influencing the occurrence and by what weight and prioritize their importance. From this prioritization process 100 can recommend or take actions that will change the outcome of the predicted event.
For example, customer relationship management (CRM) is an approach to managing a company's interaction with current and potential customers. Referring now to FIG. 4, an exemplary sales funnel progression of a CRM is illustrated. Process 100 may be used mathematically model deals that are present in each of the different phases (A to F) as well as the progression forward and backwards between phases of the sales funnel (G to M). Further, process 100 can characterize and model those deals that fail or drop out of the funnel at specific stages (N to Q) and more powerfully the process 100 can determine what is different from those deals that drop out or lose and those that progress to the next stage and eventually win. The categorical and differential delta between those that drop out and those that progress and win produce a list or opportunities for next best action to save those deals or change the approach so that they can move to the next stage in the sales funnel and how those next best actions individually can increase the chances of progression and or win per action item by probability.
As one might imagine, in addition to CRM, predictive process 100 can be used in a number of areas, such as, for example, healthcare, where it can be used in modeling real time medical management, disease progression, claims, payment integrity, and so forth.
Predictive process 100 can be used in customer facing, e.g., how to identify a customer, mathematically model who the customer is, how to contact the customer at the right time, in the right channel (e.g., email, phone, mail, app, and so forth) with the right message to drive conversion, and so forth.
Predictive process 100 can be used to acquire and retain the right customers, especially within loyalty programs.
Predictive process 100 can be used in life sciences, e.g., patient and prescriber identification.
Predictive process 100 can be used in the natural resources and oil and gas industry, e.g., how to identify what fields of natural resources are most profitable and how to develop in the most efficient and profitable manner.
Predictive process 100 can be used with commodity, stock and financial instruments, e.g. to predict price and trading value in future.
By way of one non-limiting example, use of predictive process 100 in a CRM environment is described below.
As shown in FIG. 5, an exemplary CRM framework 200 is illustrated, which depicts a process from data acquisition through four veins of analysis, i.e., structured data, supervised learning, unsupervised learning and sentiment analysis. The result is a deal win/loss probability, where the deal will fail if it does not win and what next best actions can be taken to increase probability of a win.
Data is harvested from a backend of a CRM system for both structured and unstructured data. Examples of structured data are client demographics, days in deal, days in stage, salesperson, categorization, and so forth. Structured data is anything entered in a specific header or predetermined field for the CRM basic or customizable. Unstructured data includes but not limited to free text notes in specific fields, meeting notes, follow up notes, meeting invites, acceptances and associated emails, emails pertaining to the deal, uploaded documents, uploaded presentations and any other potential files such as PDFTM files, ExcelTM files, and so forth. FIG. 6 illustrates some exemplary types of data ingested. Data is placed in a data base and then analyzed in two veins, i.e., structured and unstructured.
Structured data is aggregated and a master data file is created. Each phase or stage is mathematically modeled from horizon data. Data is then enriched from unstructured data. Machine learning is used to determine win/loss prediction classifiers. Machine learning is used to determine probability of deal movement classifiers from one stage to the next and in what direction. The Machine learning results in key deal progression drivers and situational data variables, and what next best action to take. An example of next best action is illustrated in FIG. 7 where situational data variables that are weighted in the prediction of deal event in the future and potential next best actions to improve deal win rate and progression through the sales funnel. Attribution modeling helps uncover how different intervention levers come together to drive/increase in overall deal conversion probabilities or deal TCV (directly or indirectly) at a per-deal or at a client/account level.
Unstructured data analysis typical includes three prongs, i.e., unsupervised learning, supervised learning and sentiment analysis. Unstructured data sources such as free text notes, emails and or files are harvested, the unstructured data is then parsed and extracted, data is cleaned and then vectorized. From this point the auto-prepared data is then divided into unsupervised and supervised veins of analysis. The unstructured data is then processed into clusters that are then mathematically modeled per stage and the progression between stages. The result of this process is the ability give a probability based on unstructured text and documents as to what stage in the CRM process the deal is present or transitioning. Once the different stage clusters are mathematically modeled, the unstructured data is analyzed for sentiment analysis. The result determines whether the sentiment of the document and or the emails are positive (win probability) or negative (loss probability) . Finally, unstructured data that has been cleansed, parsed and vectorized are labeled based in the unsupervised learning results of the clusters and mathematical models of each sales stage and or transitioning. After this has been achieved a machine learning neural net model is trained based on segmentation and characterization per sales stage and transition phase, which allows for auto-classification and probability referencing of unstructured data as to what phase the deal is currently assigned. The summation of the unsupervised, supervised and sentiment analysis results are fed into the data enrichment of the structured data for further machine learning modeling. The four veins of analysis, i.e., structured, unsupervised learning, supervised learning and sentiment analysis give final results of:
1. Win/Loss Probability 2. Key Deal Progression Drivers
3. Next Best Stage Prediction
4. Final Deal State Prediction
5. Realtime Deal Directionality Prediction
6. Next best sales rep
7. Next best capability
8. Deal gap analysis
9. Deal conversion and intervention/recommendation
As described above, the present invention enables predictive modeling of the progression of a customer / deal along the sales funnel by leveraging both the structured and unstructured data.
The present invention identifies the factors controlling the progression of a customer / deal along the sales funnel.
The present invention predicts the end state (target) probabilities of a new customer / deal and up to what stage (along the funnel) a new customer / a new deal can progress (if the target state probabilities are not favorable).
It would be appreciated by those skilled in the art that various changes and modifications can be made to the illustrated embodiments without departing from the spirit of the present invention. All such modifications and changes are intended to be within the scope of the present invention except as limited by the scope of the appended claims.

Claims

WHAT IS CLAIMED IS:
1. A method comprising: in a network of interconnected computer systems, receiving historical data housed in the one or more interconnected computer systems in a first computer system, the historical data comprising structured data and unstructured data; storing the historical data in a central database residing in the first computer system in the network; aggregating the historical data in the central database in the first computer system according to subject matter; validating the aggregated historical data in the first computer system; analyzing the validated aggregated historical data in the first computer system using a series of tools; generating normal profiles in the first computer system from the analyzed validated aggregated historical data; storing the generated normal profiles as horizon data sets; receiving real time data housed in the one or more of the interconnected computer systems in the first computer system; comparing the real time data to the horizon data sets to identify normal and out of normal profiles; and generating rules-based, predictive and interventive outcomes for out of normal profiles based on parameters for investigation .
2. The method of claim 1 wherein the series of tools include one or more of data science tools, analytical tools and artificial intelligence tools.
3. The method of claim 1 wherein the parameters for investigation include one or more of risk parameters, financial parameters, operations parameters, quality parameters, business parameters and industry parameters.
4. The method of claim 1 wherein receiving historical data further comprises structured and unstructured data received from additional computer systems linked to the network.
5. The method of claim 4 wherein the structured data comprises data that resides in a fixed field within a record or file.
6. The method of claim 5 wherein the unstructured data comprises data that does not have a pre-defined data model or is not organized in a pre-defined manner.
7. The method of claim 2 wherein the tools include one or more of text mining, natural language processing (NLP), natural language utilization (NLU), natural language generation (NLG), machine learning (supervised and unsupervised), genetic programming, evolutionary programming, deep learning, neural networks, computer vision and audio sound recognition.
8. The method of claim 7 wherein the normal profiles represent targeted industries.
9. The method of claim 8 wherein the normal profiles are defined by one or more mathematical patters, repeatable behaviors, characteristics, pathways, common variables, causal events, risk clusters and return on investment (ROI) clusters.
10. The method of claim 9 wherein real time data is defined as decision time data.
11. The method of claim 10 wherein decision time data comprises data that is analyzed and quantified to generate insights and predictions based on a need of an industry.
12. The method of claim 11 further comprising: introducing a normal profile or an out of normal profile into a theoretical scenario; and using the introduced normal profile or out of normal profile to predict a probable outcome of the theoretical scenario.
13. The method of claim 12 wherein using the introduced norm profile to predict a probable outcome of the theoretical scenario comprises matching mathematical profiles of the norm profile to situational event data and variables of the theoretical scenario.
14. A system comprising: a processor; a memory; a data store; the memory comprising an operating system and a predictive process, the predictive process comprising: receiving historical data housed in one or more interconnected computer systems, the historical data comprising structured data and unstructured data; storing the historical data in the data store; aggregating the historical data in the data store according to subject matter; validating the aggregated historical data; analyzing the validated aggregated historical data using a series of tools; generating normal profiles from the analyzed validated aggregated historical data; storing the generated normal profiles as horizon data sets; receiving real time data from the one or more of the interconnected computer systems; comparing the real time data to the horizon data sets to identify normal and out of normal profiles; and generating rules-based outcomes for out of normal profiles based on parameters for investigation.
15. The system of claim 14 wherein the series of tools include one or more of data science tools, analytical tools and artificial intelligence tools.
16. The system of claim 14 wherein the parameters for investigation include one or more of risk parameters, financial parameters, operations parameters, quality parameters, business parameters and industry parameters.
PCT/US2020/034561 2020-05-26 2020-05-26 Predictive and interventive intelligence WO2021242223A1 (en)

Priority Applications (5)

Application Number Priority Date Filing Date Title
MX2022014681A MX2022014681A (en) 2020-05-26 2020-05-26 Predictive and interventive intelligence.
PCT/US2020/034561 WO2021242223A1 (en) 2020-05-26 2020-05-26 Predictive and interventive intelligence
US17/999,045 US20230177537A1 (en) 2020-05-26 2020-05-26 Predictive and interventive intelliigence
EP20938380.1A EP4154209A4 (en) 2020-05-26 2020-05-26 Predictive and interventive intelligence
CN202080101992.0A CN116324832A (en) 2020-05-26 2020-05-26 Predictive intelligence

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/US2020/034561 WO2021242223A1 (en) 2020-05-26 2020-05-26 Predictive and interventive intelligence

Publications (1)

Publication Number Publication Date
WO2021242223A1 true WO2021242223A1 (en) 2021-12-02

Family

ID=78745054

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2020/034561 WO2021242223A1 (en) 2020-05-26 2020-05-26 Predictive and interventive intelligence

Country Status (5)

Country Link
US (1) US20230177537A1 (en)
EP (1) EP4154209A4 (en)
CN (1) CN116324832A (en)
MX (1) MX2022014681A (en)
WO (1) WO2021242223A1 (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070288297A1 (en) * 2006-06-07 2007-12-13 Markus Karras Industry scenario mapping tool
US20100010878A1 (en) * 2004-04-16 2010-01-14 Fortelligent, Inc. Predictive model development
US20110270646A1 (en) * 2008-07-11 2011-11-03 Gorur Narayana Srinivasa Prasanna Computer implemented decision support method & system
US20110295783A1 (en) * 2008-03-05 2011-12-01 Quantum Intelligence, Inc. Multiple Domain Anomaly Detection System and Method Using Fusion Rule and Visualization
US20180060744A1 (en) * 2014-05-23 2018-03-01 DataRobot, Inc. Systems for second-order predictive data analytics, and related methods and apparatus

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150371163A1 (en) * 2013-02-14 2015-12-24 Adaptive Spectrum And Signal Alignment, Inc. Churn prediction in a broadband network
EP3278213A4 (en) * 2015-06-05 2019-01-30 C3 IoT, Inc. Systems, methods, and devices for an enterprise internet-of-things application development platform

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100010878A1 (en) * 2004-04-16 2010-01-14 Fortelligent, Inc. Predictive model development
US20070288297A1 (en) * 2006-06-07 2007-12-13 Markus Karras Industry scenario mapping tool
US20110295783A1 (en) * 2008-03-05 2011-12-01 Quantum Intelligence, Inc. Multiple Domain Anomaly Detection System and Method Using Fusion Rule and Visualization
US20110270646A1 (en) * 2008-07-11 2011-11-03 Gorur Narayana Srinivasa Prasanna Computer implemented decision support method & system
US20180060744A1 (en) * 2014-05-23 2018-03-01 DataRobot, Inc. Systems for second-order predictive data analytics, and related methods and apparatus

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP4154209A4 *

Also Published As

Publication number Publication date
US20230177537A1 (en) 2023-06-08
CN116324832A (en) 2023-06-23
EP4154209A4 (en) 2024-01-03
EP4154209A1 (en) 2023-03-29
MX2022014681A (en) 2023-02-22

Similar Documents

Publication Publication Date Title
Chen et al. Machine learning techniques for anti-money laundering (AML) solutions in suspicious transaction detection: a review
US20240046156A1 (en) Data clean-up method for improving predictive model training
US20200111100A1 (en) Method of operating artificial intelligence machines to improve predictive model training and performance
US20190108456A1 (en) Method of automating data science services ---
US11080709B2 (en) Method of reducing financial losses in multiple payment channels upon a recognition of fraud first appearing in any one payment channel
Tida et al. Universal spam detection using transfer learning of BERT model
Raza et al. A comparative analysis of machine learning models for quality pillar assessment of SaaS services by multi-class text classification of users’ reviews
Kumar et al. An optimal churn prediction model using support vector machine with adaboost
Srinivasarao et al. Machine intelligence based hybrid classifier for spam detection and sentiment analysis of SMS messages
Kothamasu et al. Sentiment analysis on twitter data based on spider monkey optimization and deep learning for future prediction of the brands
Al Duhayyim et al. Optimized stacked autoencoder for IoT enabled financial crisis prediction model
De et al. Predicting customer churn: A systematic literature review
US20210350272A1 (en) System and method for utilizing grouped partial dependence plots and shapley additive explanations in the generation of adverse action reason codes
US10896034B2 (en) Methods and systems for automated screen display generation and configuration
Aslam et al. Improving the review classification of Google apps using combined feature embedding and deep convolutional neural network model
US20230177537A1 (en) Predictive and interventive intelliigence
US11314488B2 (en) Methods and systems for automated screen display generation and configuration
Pham et al. Fairness Aware Swarm-based Machine Learning for Data Streams
Jaiswal et al. Customer transaction prediction system
Erkayman et al. New Artificial intelligence approaches for brand switching decisions
Saeed et al. Predicting Customer’s Subscription Response to Bank Telemarketing Campaign Based on Machine learning Algorithms
El Barakaz et al. Optimization of imbalanced and multidimensional learning under bayes minimum risk and savings measure
Liu et al. A comparison of machine learning algorithms for prediction of past due service in commercial credit
Islam et al. Feature engineering based credit card fraud detection for risk minimization in e-commerce
Sam et al. Customer Churn Prediction using Machine Learning Models

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20938380

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2020938380

Country of ref document: EP

Effective date: 20221219

NENP Non-entry into the national phase

Ref country code: DE