System and method for customer journey event representation learning and outcome prediction using neural sequence models

Patent metadata

Field	Value
Publication number	US-11568305-B2
Application number	US-201916379110-A
Country	US
Kind code	B2
Filing date	Apr 9, 2019
Priority date	Apr 9, 2019
Publication date	Jan 31, 2023
Grant date	Jan 31, 2023

Abstract

A system and method are presented for customer journey event representation learning and outcome prediction using neural sequence models. A plurality of events are input into a module where each event has a schema comprising characteristics of the events and their modalities (web clicks, calls, emails, chats, etc.). The events of different modalities can be captured using different schemas and therefore embodiments described herein are schema-agnostic. Each event is represented as a vector of some number of numbers by the module with a plurality of vectors being generated in total for each customer visit. The vectors are then used in sequence learning to predict real-time next best actions or outcome probabilities in a customer journey using machine learning algorithms such as recurrent neural networks.

First claim

The invention claimed is:

1. A method for obtaining vector representations for web events comprising:
a. logging and storing events from the browsing sessions, wherein the events comprise activity data from the plurality of customers;
b. pre-processing the raw data captured on the website, wherein the pre-processing further comprises the steps of:
i. removing or masking/encrypting values in all columns containing personally identifiable information,
ii. identifying class attributes which need to be predicted when model training,
iii. identifying and removing columns which duplicate class attributes,
iv. removing array type data,
v. removing timestamp data,
vi. converting all Boolean type data into integer type, and
vii. replacing all null and unknown values, and
c. obtaining the ingestible datasets, which are capable of application to an algorithm for obtaining vector representations of web events;
d. training a feed forward neural network with the ingestible datasets;
e. inputting the web events into the feed forward neural network, wherein, the output comprises vector representations of each activity in the web events; and
f. appending vector columns to datasets for the corresponding activity in the web events,
receiving, by the feed forward neural network models, stored events data from a plurality of customer's browsing sessions on a website;
modeling, by the feed forward neural network models, a customer journey of a customer; and
providing, by the feed forward neural network models, predicted outcomes related to the customer.

2. The method of claim 1 , wherein the vector representations comprise a common n-dimensional space; and wherein the predicted outcomes comprise a sale probability during a session of the customer.

3. The method of claim 1 , wherein the training comprises prediction of the event class of input; and wherein the predicted outcomes comprise a routing of an interaction of the customer to a more preferable agent.

4. The method of claim 3 , wherein the trained feed forward neural network comprises a plurality of input layer branches.

5. The method of claim 4 , wherein the plurality of input layer branches further comprises at least a first branch accepting categorical attributes and a second branch accepting continuous numerical value attributes.

6. The method of claim 1 , wherein the events are stored with a common schema.

7. The method of claim 1 , wherein the converting of all Boolean type data into an integer type further comprises replacing ‘true’ with a 1, and ‘false’ with a 0.

8. The method of claim 1 , wherein the replacing all null and unknown values further comprises replacing all null and unknown values for integer types with a 1, with a 0.0 for double types, and ‘unknown’ for string types.

9. The method of claim 1 , wherein the pre-processing steps are performed in sequential order.

10. The method of claim 1 , wherein each event comprises a schema comprising characteristics and modalities.

11. The method of claim 1 , further comprising the steps of:
a. inputting the appended datasets into a trained neural network comprising LSTM cell units and dense neuron units, wherein input and hidden layers of the neural network comprise LSTM cell units and an output layer comprises the dense neuron units;
b. obtaining a processed sequence which is input into the dense neuron layer as a single vector;
c. applying a softmax function to the single vector; and
d. obtaining an outcome probability for the vector.

12. The method of claim 11 , wherein the neural network has been trained using data pre-processed, the method for pre-processing comprising:
a. identifying a set of outcome IDs for a given entity;
b. defining a dataframe schema capturing visit ID, time ordered sequence of event IDs within a given visit, and class label;
c. querying an event dataframe to populate the defined dataframe;
d. removing outcome events from the sequence; and
e. replacing event IDs with the respective event vectors.

13. The method of claim 11 , wherein the input layer accepts input in order of timestamp.

14. The method of claim 11 , wherein the single vector comprises a representation of the previous events in the sequence.

Assignees

• Genesys Telecommunications Laboratories Inc

Inventors

• Negi Sapna
• Dabrowski Maciej
• Ganapathiraju Aravind
• Munoz Emir
• Raghavendra Veera Elluru
• Wyss Felix Immanuel

Classifications

• G06N3/04

  Architecture, e.g. interconnection topology · CPC title

• G06N20/00 Primary

  Machine learning · CPC title

• G06F21/602

  Providing cryptographic facilities or services · CPC title

• G06Q30/0631

  Recommending goods or services · CPC title

• G06N3/08

  Learning methods · CPC title