# US 2021/0241292 A1 · System and method for profiling and predicting customer behavior

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0241292 A1
- Document ID: US20210241292A1
- Status: Published application
- Kind: A1
- Application number: 16/823,492
- Filed: 2020-03-19
- Published: 2021-08-05
- Earliest claimed: 2020-02-05
- Assignees: Fractal Analytics Private Limited
- Applicants: Fractal Analytics Private Limited
- Inventors: Amit Pandey; Yuvaneet Kumar Bhaker; Aniket Koyande; Rahul Desai
- Web page: https://patentgrep.com/patent/US20210241292A1

## CPC

- G06Q 30/0202: Market predictions or forecasting for commercial activities
- G06N 3/006: based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/045: Combinations of networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/08: Learning methods
- G06N 3/088: Non-supervised learning, e.g. competitive learning
- G06N 3/09: Supervised learning
- G06N 3/092: Reinforcement learning
- 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
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

The present disclosure describes a system and method for profiling and predicting customer behavior using machine learning. The system for profiling and predicting customer behavior uses all available signals from each individual customer across digital channels to build predictive artificial intelligence algorithms to create customer genome markers that predict the likelihood that specific customer will respond to digital actions.

## Claims

### Claim 1

A system for profiling and predicting customer behavior, comprising:
a computer server comprising instructions that when executed by one or more processors effectuate operations comprising:
retrieving, via a system database, data of a customer from a plurality of data sources for the customer;
processing, via a data harmonization engine, data fed by the system database, generating a harmonized single view per entity by standardizing features comprising entity's demographic, behaviors, needs, and preferences;
feeding, via a deep learning engine, the features into a deep learning ensemble comprising convolutional neural network (CNN), deep reinforcement learning, and recurrent neural network (RNN), and training the deep learning ensemble to generate a set of customer genome markers defining customer characteristics and suggested actions for the customer and;
providing, via a data science engine, the subset of explanatory features for every customer action using an information-theoretic predictive modeling technique based on coalition game theory;
transmitting the customer genome markers to a prediction engine to generate a recommendation for a next best action for a target consumer, wherein the prediction engine uses deep reinforcement learning to penalize or reward itself in case a prediction underperforms or beats a previous prediction cycle;
providing, via a web application and an application program interface (API), a user interface; and
enabling, via a master controller, programmable control of the system database, the data science engine, and the deep learning engine;
wherein the master controller is configured to create a schedule for scheduling data retrieving, transformation, validation, and training; and
wherein the deep learning engine collects feedbacks as to whether the customer has taken suggested actions, wherein the feedbacks are used by the deep learning ensembles to improve predictions.

### Claim 2 (depends on 1)

The system of claim 1, wherein the data are grouped by columns by a logical data model and aggregated at a level of granularity.

### Claim 3 (depends on 1)

The system of claim 1, wherein a probability score is computed for each genome marker.

### Claim 4 (depends on 1)

The system of claim 1, wherein the customer genome markers comprise a responsiveness marker, a portfolio diversification marker, a transactional diversification marker, a momentum marker, an email engagement marker, and a next business product category marker.

### Claim 5 (depends on 1)

The system of claim 1, wherein the web application provides a user interface for an admin user, a marketing user and a sales user.

### Claim 6 (depends on 5)

The system of claim 5, wherein the user interface for an admin user allows the admin user to add, delete, or update customer genome users via user settings, configure the customer genome system and the interface to external data source via system settings, view system events in real-time including access to system log files via system status, and create action segments.

### Claim 7 (depends on 5)

The system of claim 5, wherein the user interface for a marketing user comprises portfolio, action segment, mysegments and segment detail.

### Claim 8 (depends on 7)

The system of claim 7, wherein the portfolio is executed by the processors to provide analysis of key metrics of an overall portfolio covering analysis on key metrics, product categories and customer genome markers.

### Claim 9 (depends on 7)

The system of claim 7, wherein the action segment is executed by the processors to provide a list of segments defined by the admin user.

### Claim 10 (depends on 7)

The system of claim 7, wherein the mysegments is executed by the processors to create the marketing user's own marketing segments.

### Claim 11 (depends on 7)

The system of claim 7, wherein the segment detail is executed by the processors to provide a detailed analysis of a selected segment covering key metrics, profile of an average financial advisor and the financial advisor's behavioral markers.

### Claim 12 (depends on 5)

The system of claim 5, wherein the user interface for a sales user comprises a ranked order, a portfolio screen, a portfolio advisor, and find advisors.

### Claim 13 (depends on 12)

The system of claim 12, wherein the ranked order is based on a priority algorithm which is ultimately based on top/current business priorities.

### Claim 14 (depends on 12)

The system of claim 12, wherein the portfolio screen is executed by the processors to provide analysis of key metrics of the sales user's list of financial advisors in comparison to the rest of an entire sales organization.

### Claim 15 (depends on 12)

The system of claim 12, wherein the portfolio screen is executed by the processors to provide breakdown of the sales user's portfolio by product categories and analysis by each category.

### Claim 16 (depends on 12)

The system of claim 12, wherein the portfolio advisor is executed by the processors to compare and contrast the customer genome markers for the sales user's portfolio with an entire portfolio of the Financial advisors.

### Claim 17 (depends on 12)

The system of claim 12, wherein the find Advisors are executed by the processors to locate a specific cohort of financial advisors who meet one or more criteria, wherein the criteria comprise raw data points such as transactions, business parameters such as average sales, geographic information, and customer genome markers.

### Claim 18 (depends on 1)

The system of claim 1, wherein the web application is executed by the processors to load data using a component router, which delivers automatic code-splitting so users only load code required to render a view they request.

### Claim 19 (depends on 1)

The system of claim 1, wherein the API is executed by the processors to communicate with a Salesforce App.

### Claim 20 (depends on 1)

The system of claim 1, wherein data science engine comprises an exploratory data analysis (EDA) module configured to generate over 200 insights regarding underlying trends and patterns that exist in data.

### Claim 21

A method for profiling and predicting customer behavior implemented by a computer server comprising instructions that when executed, effectuate operations comprising, comprising:
retrieving, via a system database, data from a plurality of data sources for a customer;
processing, via a data harmonization engine, data fed by the system database, and generating a harmonized single view per entity by standardizing features comprising entity's demographic and behavioral traits;
selecting, via a data science engine, a subset of standardized features for every customer action using an information-theoretic predictive modeling technique based on coalition game theory;
feeding, via a deep learning engine, the subset of standardized features into a deep learning ensemble comprising convolutional neural network (CNN), deep reinforcement learning, and recurrent neural network (RNN), and training the deep learning ensemble to generate a set of customer genome markers defining customer characteristics and suggested actions for the customer;
transmitting the customer genome markers to a prediction engine to generate a recommendation for a next best action for a target consumer, wherein the prediction engine uses deep reinforcement learning to penalize or reward itself in case a prediction underperforms or beats a previous prediction cycle;
providing, via a web application and an application program interface (API), a user interface; and
enabling, via a master controller, programmable control of the system database, the data science engine, and the deep learning engine;
wherein the master controller is configured to create a schedule for scheduling data retrieving, transformation, validation, and training; and
wherein the deep learning engine collects feedbacks as to whether the customer has taken suggested actions, wherein the feedbacks are used by the deep learning ensembles to improve predictions.
