# US 12,762,580 B2 · Machine learning (ML)-based system and method for customer segmentation and worklist generation

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

- Number: US 12,762,580 B2
- Document ID: US12762580B2
- Status: Granted
- Kind: B2
- Application number: 18/306,281
- Filed: 2023-04-25
- Published: 2023-10-26
- Granted: 2026-10-13
- Earliest claimed: 2022-04-25
- Assignees: Highradius Corporation
- Applicants: Highradius Corporation
- Inventors: Anupam Kunwar; Glen George; Apoorva Shrivastava; Sayanta Mukherjee; Shailly Shailja
- Web page: https://patentgrep.com/patent/US12762580B2

## CPC

- G06Q 30/0201: Market modelling; Market analysis; Collecting market data
- G06Q 10/0635: Risk analysis of enterprise or organisation activities
- G06Q 20/102: {Bill distribution or payments}
- G06Q 30/0204: Market segmentation
- G06Q 30/04: Billing or invoicing
- G06Q 40/03: Credit; Loans; Processing thereof
- 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

## Abstract

A Machine Learning (ML)-based computing system and method for customer segmentation and collections worklist generation is disclosed. The method includes data receiver module configured to receive one or more open invoices from one or more electronic devices associated with a user. Further, the method includes data processing module configured to process the received one or more open invoices. Further, the method includes segment clustering module configured to determine a payment behaviour and customer risk. Further, the method includes worklist generation module configured to generate a customer worklist. Further, the method includes action recommendation module configured to recommend one or more collection strategies. Further, the method includes prioritization module configured to rank each of the one or more customers in the generated worklist. Furthermore, the method includes the data output module configured to output the one or more collection strategies and the ranked one or more customers.

## Claims

### Claim 1

A Machine Learning (ML)-based computing method for customer segmentation and collections worklist generation, the ML-based computing method comprising:
receiving, by one or more hardware processors, one or more open invoices from one or more electronic devices associated with a user, wherein the invoice data comprises customer number, payment terms, invoice date, baseline date, due date, discount date, clearing date, posting date, invoice type, transactions within a company, invoice amount, discount amount and payment block status;
processing, by the one or more hardware processors, the received one or more open invoices to generate granularity level instances, granularity level detected pattern, payment pattern consistency score and predicted payment date of one or more open invoices based on a trained ML model, wherein processing the received one or more open invoices comprises:
determining, by the one or more hardware processors, an adherence evaluation window and a look-back period for each date-shift-payment-pattern feature corresponding to each set of granularity level instances based on historically closed invoice data;
estimating, by the one or more hardware processors, a cumulative error distribution of historical deviations between predicted and actual clearing dates of invoices within the look-back period;
determining, by the one or more hardware processors, from the cumulative error distribution, a revenue concentration level within the adherence evaluation window; and
identifying, by the one or more hardware processors, the date-shift-payment-pattern feature having the highest probability of adherence as an optimal payment pattern feature based on the determined revenue concentration level;
predicting, by the one or more hardware processors, a financial transaction completion pattern for the set of open invoices based on the optimal payment pattern feature, the set of date-shift-payment-pattern features and the set of granularity level instances by using a financial transaction pattern prediction based ML model; and
predicting, by the one or more hardware processors, the payment date for the one or more open invoices and the payment pattern consistency scores based on the optimal payment pattern features and the financial transaction completion pattern;
determining, by the one or more hardware processors, a payment behavior and customer risk for each of the one or more customers using a clustering based ML model, wherein the clustering based ML model groups similar data points associated with the generated payment pattern consistency scores, delay propensity factor, current state defining factors of the invoice and customer, contact behavior based features of customer, granularity level detected pattern and predicted payment date into clusters based on at least one of: distances between each of the data points, density of the data points, hierarchy of the clusters, and similarity to one or more representative points, wherein the clustering based ML model comprises at least one of: a K-means model, DBSCAN model, HDBSCAN model and K-medoids model, and wherein the clustering based ML model is automatically selected for datasets having at least one of: outliers and varying cluster densities to enable at least one of: a noise-resistant and hierarchy-based clustering;
segmenting, by the one or more hardware processors, each of the granularity level instances dynamically into one or more categories based on the determined payment behavior and the customer risk using the clustering based ML model, wherein the one or more categories comprises payment behaviour based customer segment and risk customer segment;
generating, by the one or more hardware processors, a customer worklist based on a configured optimization metric and the one or more categories, wherein the configured optimization metric comprises working capital, collecting the payment for the one or more open invoices;
recommending, by the one or more hardware processors, one or more collection strategies for each of the one or more categories based on type of a category, a severity, the configured optimization metric and the generated customer worklist, wherein the one or more collection strategies comprises one or more actions;
ranking, by the one or more hardware processors, each of the one or more customers in the generated customer worklist based on the one or more recommended collection strategies;
dynamically updating and retraining, by the one or more hardware processors, the trained ML model by:
monitoring performance of the trained ML model in real time;
identifying instances where performance of the ML-based computing system falls below a predetermined threshold;
automatically generating new training data based on the identified instances;
retraining the trained ML model using the newly generated training data; and
outputting, by the one or more hardware processors, the one or more collection strategies and the ranked one or more customers on a user interface of one or more electronic devices associated with the user.

### Claim 2 (depends on 1)

The ML-based computing method of in claim 1, wherein processing the received one or more open invoices to generate the granularity level detected pattern, payment pattern consistency score and predicted payment date of the one or more open invoices based on the trained ML model, the method comprising:
extracting a set of granularity levels from the received invoice data based on data density, volume and the set of data attributes, wherein the set of granularity levels comprises at least one of company code, customer number and payment terms;
generating a set of granularity level instances based on the extracted set of granularity levels, wherein each of the set of granularity level instances corresponds to a basic logical unit for which payment transaction pattern is to be determined;
creating the generated set of granularity level instances based on a model configuration for granularity levels, wherein the model configuration is used to generate the set of granularity level instances;
computing a payment frequency bucket feature for the created set of granularity level instances;
assigning the one or more open invoices to clusters based on the computed payment frequency bucket feature and associated with the data density of the extracted set of granularity levels; and
generating the payment pattern features by shifting invoice level data attributes to plurality of reference points on a calendar based on the computed payment frequency bucket feature of the cluster.

### Claim 3 (depends on 1)

The ML-based computing method of claim 1, wherein determining the payment behavior and customer risk for each of the one or more customers based on the payment pattern consistency scores, the delay propensity factor, current state defining factors of the invoice and customer, contact behavior based features of the customer, granularity level detected pattern and predicted payment date, the method comprising:
generating the payment pattern consistency scores for the detected payment pattern features;
calculating the delay propensity factor of the selected payment pattern from a due date of the one or more open invoices;
extracting the current state defining factors from the received one or more open invoices, wherein the determined current state defining factors of the invoice and customer comprise previous payment week of the customer, previous posting week of the customer, current open amount of the customer, total post due amount of the customer, payment frequency factor, skipped invoice flag and disputed flag; and
extracting the contact behavior based features of the customer, wherein the determined contact behavior based features of the customer comprises contact date and types, email responsiveness, phone and email preference, number of communication in recent time period.

### Claim 4 (depends on 1)

The ML-based computing method of claim 1, wherein the payment behaviour based customer segment comprises early payers, pays within one week from the due date, pays after one week from the due date and less predictable customers and wherein the risk customer segment comprises growing past due amount, uncleared invoices and termination of monetary transactions.

### Claim 5 (depends on 1)

The ML-based computing method of claim 1, wherein the one or more actions comprises proactive follow-up, active follow-up and relationship building call, email reminder, sending statement of account, dunning notices and escalate to internal stakeholders.

### Claim 6 (depends on 1)

The ML-based computing method of claim 1, wherein ranking each of the one or more customers in the generated customer worklist based on at least one of: the one or more recommended collection strategies, the method comprising:
generating a summary view of the payment behaviour based customer segment, wherein the summary view of the each payment behaviour based customer segment comprises total outstanding amount, current amount, past due amount, Days Sale Outstanding (DSO), BPDSO, target for current month, total due overview, total due analysis and DSO impacts.

### Claim 7

A Machine Learning (ML)-based computing system for processing of open invoices for the customers, the ML-based computing system comprising:
one or more hardware processors; and
a memory coupled to the one or more hardware processors, wherein the memory comprises a plurality of modules in the form of programmable instructions executable by the one or more hardware processors, and wherein the plurality of modules comprises:
a data receiver module configured to receive one or more open invoices from one or more electronic devices associated with a user, wherein the invoice data comprises customer number, payment terms, invoice date, baseline date, due date, discount date, clearing date, posting date, invoice type, transactions within a company, invoice amount, discount amount and payment block status;
a data processing module configured to process the received one or more open invoices to generate granularity level instances, granularity level detected pattern, payment pattern consistency score and predicted payment date of one or more open invoices based on a trained ML model,
wherein to process the received one or more open invoices, a payment pattern selection module is configured to:
determine an adherence evaluation window and a look-back period for each date-shift-payment-pattern feature corresponding to each set of granularity level instances based on historically closed invoice data;
estimate a cumulative error distribution of historical deviations between predicted and actual clearing dates of invoices within the look-back period;
determine, from the cumulative error distribution, a revenue concentration level within the adherence evaluation window; and
identify the date-shift-payment-pattern feature having the highest probability of adherence as an optimal payment pattern feature based on the determined revenue concentration level;
a data prediction module configured to:
predict a financial transaction completion pattern for the set of open invoices based on the optimal payment pattern feature, the set of date-shift-payment-pattern features and the set of granularity level instances by using a financial transaction pattern prediction-based ML model;
predict the payment date for the one or more open invoices and the payment pattern consistency scores based on the optimal payment pattern features and the financial transaction completion pattern;
a segment clustering module configured to:
determine a payment behavior and customer risk for each of the one or more customers using a clustering based ML model, wherein the clustering based ML model groups similar data points associated with the generated payment pattern consistency scores, delay propensity factor, current state defining factors of the invoice and customer, contact behavior based features of customer, granularity level detected pattern and predicted payment date into clusters based on at least one of: distances between each of the data points, density of the data points, hierarchy of the clusters, and similarity to one or more representative points, and wherein the clustering based ML model comprises at least one of: a K-means model, DBSCAN model, HDBSCAN model and K-medoids model, and wherein the clustering based ML model is automatically selected for datasets having at least one of: outliers and varying cluster densities to enable at least one of: a noise-resistant and hierarchy-based clustering;
segment each of the granularity level instances dynamically into one or more categories based on the determined payment behavior and the customer risk using the clustering based ML model, wherein the one or more categories comprises payment behaviour based segment and high risk segments;
a worklist generation module configured to generate a customer worklist based on a configured optimization metric and the one or more categories, wherein the configured optimization metric comprises working capital, collecting the payment for the one or more open invoices;
an action recommendation module configured to recommend one or more collection strategies for each of the one or more categories based on type of a category, a severity, and the configured optimization metric, wherein the one or more collection strategies comprises one or more actions;
a prioritization module configured to rank each of the one or more customers in the generated customer worklist based on at least one of: the one or more recommended collection strategies;
wherein the ML-based computing system is configured to:
monitor performance of the trained ML model in real time;
identify instances where performance of the ML-based computing system falls below a predetermined threshold;
automatically generate new training data based on the identified instances; and
retrain the trained ML model using the newly generated training data;
a data output module configured to output the one or more collection strategies and the ranked one or more customers on a user interface of one or more electronic devices associated with the user.

### Claim 8 (depends on 7)

The ML-based computing system of claim 7, wherein to process the received one or more open invoices to generate the granularity level detected pattern, payment pattern consistency score and predicted payment date of the one or more open invoices based on the trained ML model, the data processing module is configured to:
extract a set of granularity levels from the received invoice data based on data density, volume and the set of data attributes, wherein the set of granularity levels comprises at least one of company code, customer number and payment terms;
generate a set of granularity level instances based on the extracted set of granularity levels, wherein each of the set of granularity level instances corresponds to a basic logical unit for which payment transaction pattern is to be determined;
create the generated set of granularity level instances based on a model configuration for granularity levels, wherein the model configuration is used to generate the set of granularity level instances;
compute a payment frequency bucket feature for the created set of granularity level instances;
assign the one or more open invoices to clusters based on the computed payment frequency bucket feature and associated with the data density of the extracted set of granularity levels;
generate the payment pattern features by shifting invoice level data attributes to plurality of reference points on a calendar based on the computed payment frequency bucket feature of the cluster;
select the payment pattern feature with highest probability of adherence for each set of granularity level instances based on the generated payment pattern features; and
predict a payment date for the one or more open invoices based on the selected payment pattern features, wherein the payment date is predicted for the one or more open invoices based on the detected payment patterns, wherein the detected payment pattern features for the set of granularity level instances facilitates the detectable payment pattern features along with the payment pattern consistency scores.

### Claim 9 (depends on 7)

The ML-based computing system of claim 7, wherein to determine the payment behavior and customer risk for each of the one or more customers based on the payment pattern consistency scores, the delay propensity factor, current state defining factors of the invoice and customer, contact behavior based features of the customer, granularity level detected pattern and predicted payment date, the segment clustering module comprising:
generate the payment pattern consistency scores for the detected payment pattern features;
calculate the delay propensity factor of the selected payment pattern from a due date of the one or more open invoices;
extract the current state defining factors from the received one or more open invoices, wherein the determined current state defining factors of the invoice and customer comprise previous payment week of the customer, previous posting week of the customer, current open amount of the customer, total post due amount of the customer, payment frequency factor, skipped invoice flag and disputed flag; and
extract the contact behavior based features of the customer, wherein the determined contact behavior based features of the customer comprises contact date and types, email responsiveness, phone and email preference, number of communication in recent time period.

### Claim 10 (depends on 7)

The ML-based computing system of claim 7, wherein the payment behaviour based customer segment comprises early payers, pays within one week from the due date, pays after one week from the due date and less predictable customers and wherein the risk customer segment comprises growing past due amount, uncleared invoices and termination of monetary transactions.

### Claim 11 (depends on 7)

The ML-based computing system of claim 7, wherein the one or more actions comprises proactive follow-up, active follow-up and relationship building call, email reminder, sending statement of account, dunning notices and escalate to internal stakeholders.

### Claim 12 (depends on 7)

The ML-based computing system of claim 7, wherein in rank each of the one or more customers in the generated customer worklist based on at least one of: the one or more recommended collection strategies, the prioritization module is configured to:
generate a summary view of the payment behaviour based customer segment, wherein the summary view of the each payment behaviour based customer segment comprises total outstanding amount, current amount, past due amount, Days Sale Outstanding (DSO), BPDSO, target for current month, total due overview, total due analysis and DSO impacts.

### Claim 13

A non-transitory computer-readable storage medium having instructions stored therein that, when executed by a hardware processor, cause the processor to perform method steps comprising:
receiving one or more open invoices from one or more electronic devices associated with a user, wherein the invoice data comprises customer number, payment terms, invoice date, baseline date, due date, discount date, clearing date, posting date, invoice type, transactions within a company, invoice amount, discount amount and payment block status;
processing the received one or more open invoices to generate granularity level instances, granularity level detected pattern, payment pattern consistency score and predicted payment date of one or more open invoices based on a trained ML model, wherein processing the received one or more open invoices comprises:
determining, by the one or more hardware processors, an adherence evaluation window and a look-back period for each date-shift-payment-pattern feature corresponding to each set of granularity level instances based on historically closed invoice data;
estimating, by the one or more hardware processors, a cumulative error distribution of historical deviations between predicted and actual clearing dates of invoices within the look-back period;
determining, by the one or more hardware processors, from the cumulative error distribution, a revenue concentration level within the adherence evaluation window; and
identifying, by the one or more hardware processors, the date-shift-payment-pattern feature having the highest probability of adherence as an optimal payment pattern feature based on the determined revenue concentration level;
predicting, by the one or more hardware processors, a financial transaction completion pattern for the set of open invoices based on the optimal payment pattern feature, the set of date-shift-payment-pattern features and the set of granularity level instances by using a financial transaction pattern prediction based ML model; and
predicting, by the one or more hardware processors, the payment date for the one or more open invoices and the payment pattern consistency scores based on the optimal payment pattern features and the financial transaction completion pattern;
determining a payment behavior and customer risk for each of the one or more customers using a clustering based ML model, wherein the clustering based ML model groups similar data points associated with the generated payment pattern consistency scores, delay propensity factor, current state defining factors of the invoice and customer, contact behavior based features of customer, granularity level detected pattern and predicted payment date into clusters based on at least one of: distances between each of the data points, density of the data points, hierarchy of the clusters, and similarity to one or more representative points, and wherein the clustering based ML model comprises at least one of: a K-means model, DBSCAN model, HDBSCAN model and K-medoids model, and wherein the clustering based ML model is automatically selected for datasets having at least one of: outliers and varying cluster densities to enable at least one of: a noise-resistant and hierarchy-based clustering;
segmenting each of the granularity level instances dynamically into one or more categories based on the determined payment behavior and the customer risk using the clustering based ML model, wherein the one or more categories comprises payment behaviour based customer segment and risk customer segment;
generating a customer worklist based on a configured optimization metric and the one or more categories, wherein the configured optimization metric comprises working capital, collecting the payment for the one or more open invoices;
recommending one or more collection strategies for each of the one or more categories based on type of a category, a severity, the configured optimization metric and the generated customer worklist, wherein the one or more collection strategies comprises one or more actions;
ranking each of the one or more customers in the generated customer worklist based on at least one of: the one or more recommended collection strategies;
dynamically updating and retraining, by the one or more hardware processors, the trained ML model by:
monitoring performance of the trained ML model in real time, identifying instances where performance of the ML-based computing system falls below a predetermined threshold;
automatically generating new training data based on the identified instances and retraining the trained ML model using the new training data; and
outputting the one or more collection strategies and the ranked one or more customers on a user interface of one or more electronic devices associated with the user.

### Claim 14 (depends on 13)

The non-transitory computer-readable storage medium of claim 13, further comprising instructions to cause the processor to perform:
extracting a set of granularity levels from the received invoice data based on data density, volume and the set of data attributes, wherein the set of granularity levels comprises at least one of company code, customer number and payment terms;
generating a set of granularity level instances based on the extracted set of granularity levels, wherein each of the set of granularity level instances corresponds to a basic logical unit for which payment transaction pattern is to be determined;
creating the generated set of granularity level instances based on a model configuration for granularity levels, wherein the model configuration is used to generate the set of granularity level instances;
computing a payment frequency bucket feature for the created set of granularity level instances;
assigning the one or more open invoices to clusters based on the computed payment frequency bucket feature and associated with the data density of the extracted set of granularity levels;
generating the payment pattern features by shifting invoice level data attributes to plurality of reference points on a calendar based on the computed payment frequency bucket feature of the cluster;
selecting the payment pattern feature with highest probability of adherence for each set of granularity level instances based on the generated payment pattern features; and
predicting a payment date for the one or more open invoices based on the selected payment pattern features, wherein the payment date is predicted for the one or more open invoices based on the detected payment patterns, wherein the detected payment pattern features for the set of granularity level instances facilitates the detectable payment pattern features along with the payment pattern consistency scores.